Upload folder using huggingface_hub

.gitattributes

@@ -43,3 +43,24 @@ sd_feed/assets/pinterest.png filter=lfs diff=lfs merge=lfs -text
 sd-3dmodel-loader/models/Samba[[:space:]]Dancing.fbx filter=lfs diff=lfs merge=lfs -text
 sd-3dmodel-loader/models/pose.vrm filter=lfs diff=lfs merge=lfs -text
 sd-webui-3d-open-pose-editor/downloads/pose/0.5.1675469404/pose_solution_packed_assets.data filter=lfs diff=lfs merge=lfs -text
+SD-CN-Animation/examples/flower_1.gif filter=lfs diff=lfs merge=lfs -text
+SD-CN-Animation/examples/flower_1.mp4 filter=lfs diff=lfs merge=lfs -text
+SD-CN-Animation/examples/flower_11.mp4 filter=lfs diff=lfs merge=lfs -text
+SD-CN-Animation/examples/girl_org.gif filter=lfs diff=lfs merge=lfs -text
+SD-CN-Animation/examples/girl_to_jc.gif filter=lfs diff=lfs merge=lfs -text
+SD-CN-Animation/examples/girl_to_jc.mp4 filter=lfs diff=lfs merge=lfs -text
+SD-CN-Animation/examples/girl_to_wc.gif filter=lfs diff=lfs merge=lfs -text
+SD-CN-Animation/examples/girl_to_wc.mp4 filter=lfs diff=lfs merge=lfs -text
+SD-CN-Animation/examples/gold_1.gif filter=lfs diff=lfs merge=lfs -text
+SD-CN-Animation/examples/macaroni_1.gif filter=lfs diff=lfs merge=lfs -text
+SD-CN-Animation/examples/tree_2.gif filter=lfs diff=lfs merge=lfs -text
+SD-CN-Animation/examples/tree_2.mp4 filter=lfs diff=lfs merge=lfs -text
+ebsynth_utility/imgs/sample1.mp4 filter=lfs diff=lfs merge=lfs -text
+ebsynth_utility/imgs/sample2.mp4 filter=lfs diff=lfs merge=lfs -text
+ebsynth_utility/imgs/sample3.mp4 filter=lfs diff=lfs merge=lfs -text
+ebsynth_utility/imgs/sample4.mp4 filter=lfs diff=lfs merge=lfs -text
+ebsynth_utility/imgs/sample5.mp4 filter=lfs diff=lfs merge=lfs -text
+ebsynth_utility/imgs/sample6.mp4 filter=lfs diff=lfs merge=lfs -text
+ebsynth_utility/imgs/sample_anyaheh.mp4 filter=lfs diff=lfs merge=lfs -text
+ebsynth_utility/imgs/sample_autotag.mp4 filter=lfs diff=lfs merge=lfs -text
+ebsynth_utility/imgs/sample_clipseg.mp4 filter=lfs diff=lfs merge=lfs -text

Abysz-LAB-Ext/LICENSE

@@ -0,0 +1,674 @@
+                    GNU GENERAL PUBLIC LICENSE
+                       Version 3, 29 June 2007
+
+ Copyright (C) 2007 Free Software Foundation, Inc. <https://fsf.org/>
+ Everyone is permitted to copy and distribute verbatim copies
+ of this license document, but changing it is not allowed.
+
+                            Preamble
+
+  The GNU General Public License is a free, copyleft license for
+software and other kinds of works.
+
+  The licenses for most software and other practical works are designed
+to take away your freedom to share and change the works.  By contrast,
+the GNU General Public License is intended to guarantee your freedom to
+share and change all versions of a program--to make sure it remains free
+software for all its users.  We, the Free Software Foundation, use the
+GNU General Public License for most of our software; it applies also to
+any other work released this way by its authors.  You can apply it to
+your programs, too.
+
+  When we speak of free software, we are referring to freedom, not
+price.  Our General Public Licenses are designed to make sure that you
+have the freedom to distribute copies of free software (and charge for
+them if you wish), that you receive source code or can get it if you
+want it, that you can change the software or use pieces of it in new
+free programs, and that you know you can do these things.
+
+  To protect your rights, we need to prevent others from denying you
+these rights or asking you to surrender the rights.  Therefore, you have
+certain responsibilities if you distribute copies of the software, or if
+you modify it: responsibilities to respect the freedom of others.
+
+  For example, if you distribute copies of such a program, whether
+gratis or for a fee, you must pass on to the recipients the same
+freedoms that you received.  You must make sure that they, too, receive
+or can get the source code.  And you must show them these terms so they
+know their rights.
+
+  Developers that use the GNU GPL protect your rights with two steps:
+(1) assert copyright on the software, and (2) offer you this License
+giving you legal permission to copy, distribute and/or modify it.
+
+  For the developers' and authors' protection, the GPL clearly explains
+that there is no warranty for this free software.  For both users' and
+authors' sake, the GPL requires that modified versions be marked as
+changed, so that their problems will not be attributed erroneously to
+authors of previous versions.
+
+  Some devices are designed to deny users access to install or run
+modified versions of the software inside them, although the manufacturer
+can do so.  This is fundamentally incompatible with the aim of
+protecting users' freedom to change the software.  The systematic
+pattern of such abuse occurs in the area of products for individuals to
+use, which is precisely where it is most unacceptable.  Therefore, we
+have designed this version of the GPL to prohibit the practice for those
+products.  If such problems arise substantially in other domains, we
+stand ready to extend this provision to those domains in future versions
+of the GPL, as needed to protect the freedom of users.
+
+  Finally, every program is threatened constantly by software patents.
+States should not allow patents to restrict development and use of
+software on general-purpose computers, but in those that do, we wish to
+avoid the special danger that patents applied to a free program could
+make it effectively proprietary.  To prevent this, the GPL assures that
+patents cannot be used to render the program non-free.
+
+  The precise terms and conditions for copying, distribution and
+modification follow.
+
+                       TERMS AND CONDITIONS
+
+  0. Definitions.
+
+  "This License" refers to version 3 of the GNU General Public License.
+
+  "Copyright" also means copyright-like laws that apply to other kinds of
+works, such as semiconductor masks.
+
+  "The Program" refers to any copyrightable work licensed under this
+License.  Each licensee is addressed as "you".  "Licensees" and
+"recipients" may be individuals or organizations.
+
+  To "modify" a work means to copy from or adapt all or part of the work
+in a fashion requiring copyright permission, other than the making of an
+exact copy.  The resulting work is called a "modified version" of the
+earlier work or a work "based on" the earlier work.
+
+  A "covered work" means either the unmodified Program or a work based
+on the Program.
+
+  To "propagate" a work means to do anything with it that, without
+permission, would make you directly or secondarily liable for
+infringement under applicable copyright law, except executing it on a
+computer or modifying a private copy.  Propagation includes copying,
+distribution (with or without modification), making available to the
+public, and in some countries other activities as well.
+
+  To "convey" a work means any kind of propagation that enables other
+parties to make or receive copies.  Mere interaction with a user through
+a computer network, with no transfer of a copy, is not conveying.
+
+  An interactive user interface displays "Appropriate Legal Notices"
+to the extent that it includes a convenient and prominently visible
+feature that (1) displays an appropriate copyright notice, and (2)
+tells the user that there is no warranty for the work (except to the
+extent that warranties are provided), that licensees may convey the
+work under this License, and how to view a copy of this License.  If
+the interface presents a list of user commands or options, such as a
+menu, a prominent item in the list meets this criterion.
+
+  1. Source Code.
+
+  The "source code" for a work means the preferred form of the work
+for making modifications to it.  "Object code" means any non-source
+form of a work.
+
+  A "Standard Interface" means an interface that either is an official
+standard defined by a recognized standards body, or, in the case of
+interfaces specified for a particular programming language, one that
+is widely used among developers working in that language.
+
+  The "System Libraries" of an executable work include anything, other
+than the work as a whole, that (a) is included in the normal form of
+packaging a Major Component, but which is not part of that Major
+Component, and (b) serves only to enable use of the work with that
+Major Component, or to implement a Standard Interface for which an
+implementation is available to the public in source code form.  A
+"Major Component", in this context, means a major essential component
+(kernel, window system, and so on) of the specific operating system
+(if any) on which the executable work runs, or a compiler used to
+produce the work, or an object code interpreter used to run it.
+
+  The "Corresponding Source" for a work in object code form means all
+the source code needed to generate, install, and (for an executable
+work) run the object code and to modify the work, including scripts to
+control those activities.  However, it does not include the work's
+System Libraries, or general-purpose tools or generally available free
+programs which are used unmodified in performing those activities but
+which are not part of the work.  For example, Corresponding Source
+includes interface definition files associated with source files for
+the work, and the source code for shared libraries and dynamically
+linked subprograms that the work is specifically designed to require,
+such as by intimate data communication or control flow between those
+subprograms and other parts of the work.
+
+  The Corresponding Source need not include anything that users
+can regenerate automatically from other parts of the Corresponding
+Source.
+
+  The Corresponding Source for a work in source code form is that
+same work.
+
+  2. Basic Permissions.
+
+  All rights granted under this License are granted for the term of
+copyright on the Program, and are irrevocable provided the stated
+conditions are met.  This License explicitly affirms your unlimited
+permission to run the unmodified Program.  The output from running a
+covered work is covered by this License only if the output, given its
+content, constitutes a covered work.  This License acknowledges your
+rights of fair use or other equivalent, as provided by copyright law.
+
+  You may make, run and propagate covered works that you do not
+convey, without conditions so long as your license otherwise remains
+in force.  You may convey covered works to others for the sole purpose
+of having them make modifications exclusively for you, or provide you
+with facilities for running those works, provided that you comply with
+the terms of this License in conveying all material for which you do
+not control copyright.  Those thus making or running the covered works
+for you must do so exclusively on your behalf, under your direction
+and control, on terms that prohibit them from making any copies of
+your copyrighted material outside their relationship with you.
+
+  Conveying under any other circumstances is permitted solely under
+the conditions stated below.  Sublicensing is not allowed; section 10
+makes it unnecessary.
+
+  3. Protecting Users' Legal Rights From Anti-Circumvention Law.
+
+  No covered work shall be deemed part of an effective technological
+measure under any applicable law fulfilling obligations under article
+11 of the WIPO copyright treaty adopted on 20 December 1996, or
+similar laws prohibiting or restricting circumvention of such
+measures.
+
+  When you convey a covered work, you waive any legal power to forbid
+circumvention of technological measures to the extent such circumvention
+is effected by exercising rights under this License with respect to
+the covered work, and you disclaim any intention to limit operation or
+modification of the work as a means of enforcing, against the work's
+users, your or third parties' legal rights to forbid circumvention of
+technological measures.
+
+  4. Conveying Verbatim Copies.
+
+  You may convey verbatim copies of the Program's source code as you
+receive it, in any medium, provided that you conspicuously and
+appropriately publish on each copy an appropriate copyright notice;
+keep intact all notices stating that this License and any
+non-permissive terms added in accord with section 7 apply to the code;
+keep intact all notices of the absence of any warranty; and give all
+recipients a copy of this License along with the Program.
+
+  You may charge any price or no price for each copy that you convey,
+and you may offer support or warranty protection for a fee.
+
+  5. Conveying Modified Source Versions.
+
+  You may convey a work based on the Program, or the modifications to
+produce it from the Program, in the form of source code under the
+terms of section 4, provided that you also meet all of these conditions:
+
+    a) The work must carry prominent notices stating that you modified
+    it, and giving a relevant date.
+
+    b) The work must carry prominent notices stating that it is
+    released under this License and any conditions added under section
+    7.  This requirement modifies the requirement in section 4 to
+    "keep intact all notices".
+
+    c) You must license the entire work, as a whole, under this
+    License to anyone who comes into possession of a copy.  This
+    License will therefore apply, along with any applicable section 7
+    additional terms, to the whole of the work, and all its parts,
+    regardless of how they are packaged.  This License gives no
+    permission to license the work in any other way, but it does not
+    invalidate such permission if you have separately received it.
+
+    d) If the work has interactive user interfaces, each must display
+    Appropriate Legal Notices; however, if the Program has interactive
+    interfaces that do not display Appropriate Legal Notices, your
+    work need not make them do so.
+
+  A compilation of a covered work with other separate and independent
+works, which are not by their nature extensions of the covered work,
+and which are not combined with it such as to form a larger program,
+in or on a volume of a storage or distribution medium, is called an
+"aggregate" if the compilation and its resulting copyright are not
+used to limit the access or legal rights of the compilation's users
+beyond what the individual works permit.  Inclusion of a covered work
+in an aggregate does not cause this License to apply to the other
+parts of the aggregate.
+
+  6. Conveying Non-Source Forms.
+
+  You may convey a covered work in object code form under the terms
+of sections 4 and 5, provided that you also convey the
+machine-readable Corresponding Source under the terms of this License,
+in one of these ways:
+
+    a) Convey the object code in, or embodied in, a physical product
+    (including a physical distribution medium), accompanied by the
+    Corresponding Source fixed on a durable physical medium
+    customarily used for software interchange.
+
+    b) Convey the object code in, or embodied in, a physical product
+    (including a physical distribution medium), accompanied by a
+    written offer, valid for at least three years and valid for as
+    long as you offer spare parts or customer support for that product
+    model, to give anyone who possesses the object code either (1) a
+    copy of the Corresponding Source for all the software in the
+    product that is covered by this License, on a durable physical
+    medium customarily used for software interchange, for a price no
+    more than your reasonable cost of physically performing this
+    conveying of source, or (2) access to copy the
+    Corresponding Source from a network server at no charge.
+
+    c) Convey individual copies of the object code with a copy of the
+    written offer to provide the Corresponding Source.  This
+    alternative is allowed only occasionally and noncommercially, and
+    only if you received the object code with such an offer, in accord
+    with subsection 6b.
+
+    d) Convey the object code by offering access from a designated
+    place (gratis or for a charge), and offer equivalent access to the
+    Corresponding Source in the same way through the same place at no
+    further charge.  You need not require recipients to copy the
+    Corresponding Source along with the object code.  If the place to
+    copy the object code is a network server, the Corresponding Source
+    may be on a different server (operated by you or a third party)
+    that supports equivalent copying facilities, provided you maintain
+    clear directions next to the object code saying where to find the
+    Corresponding Source.  Regardless of what server hosts the
+    Corresponding Source, you remain obligated to ensure that it is
+    available for as long as needed to satisfy these requirements.
+
+    e) Convey the object code using peer-to-peer transmission, provided
+    you inform other peers where the object code and Corresponding
+    Source of the work are being offered to the general public at no
+    charge under subsection 6d.
+
+  A separable portion of the object code, whose source code is excluded
+from the Corresponding Source as a System Library, need not be
+included in conveying the object code work.
+
+  A "User Product" is either (1) a "consumer product", which means any
+tangible personal property which is normally used for personal, family,
+or household purposes, or (2) anything designed or sold for incorporation
+into a dwelling.  In determining whether a product is a consumer product,
+doubtful cases shall be resolved in favor of coverage.  For a particular
+product received by a particular user, "normally used" refers to a
+typical or common use of that class of product, regardless of the status
+of the particular user or of the way in which the particular user
+actually uses, or expects or is expected to use, the product.  A product
+is a consumer product regardless of whether the product has substantial
+commercial, industrial or non-consumer uses, unless such uses represent
+the only significant mode of use of the product.
+
+  "Installation Information" for a User Product means any methods,
+procedures, authorization keys, or other information required to install
+and execute modified versions of a covered work in that User Product from
+a modified version of its Corresponding Source.  The information must
+suffice to ensure that the continued functioning of the modified object
+code is in no case prevented or interfered with solely because
+modification has been made.
+
+  If you convey an object code work under this section in, or with, or
+specifically for use in, a User Product, and the conveying occurs as
+part of a transaction in which the right of possession and use of the
+User Product is transferred to the recipient in perpetuity or for a
+fixed term (regardless of how the transaction is characterized), the
+Corresponding Source conveyed under this section must be accompanied
+by the Installation Information.  But this requirement does not apply
+if neither you nor any third party retains the ability to install
+modified object code on the User Product (for example, the work has
+been installed in ROM).
+
+  The requirement to provide Installation Information does not include a
+requirement to continue to provide support service, warranty, or updates
+for a work that has been modified or installed by the recipient, or for
+the User Product in which it has been modified or installed.  Access to a
+network may be denied when the modification itself materially and
+adversely affects the operation of the network or violates the rules and
+protocols for communication across the network.
+
+  Corresponding Source conveyed, and Installation Information provided,
+in accord with this section must be in a format that is publicly
+documented (and with an implementation available to the public in
+source code form), and must require no special password or key for
+unpacking, reading or copying.
+
+  7. Additional Terms.
+
+  "Additional permissions" are terms that supplement the terms of this
+License by making exceptions from one or more of its conditions.
+Additional permissions that are applicable to the entire Program shall
+be treated as though they were included in this License, to the extent
+that they are valid under applicable law.  If additional permissions
+apply only to part of the Program, that part may be used separately
+under those permissions, but the entire Program remains governed by
+this License without regard to the additional permissions.
+
+  When you convey a copy of a covered work, you may at your option
+remove any additional permissions from that copy, or from any part of
+it.  (Additional permissions may be written to require their own
+removal in certain cases when you modify the work.)  You may place
+additional permissions on material, added by you to a covered work,
+for which you have or can give appropriate copyright permission.
+
+  Notwithstanding any other provision of this License, for material you
+add to a covered work, you may (if authorized by the copyright holders of
+that material) supplement the terms of this License with terms:
+
+    a) Disclaiming warranty or limiting liability differently from the
+    terms of sections 15 and 16 of this License; or
+
+    b) Requiring preservation of specified reasonable legal notices or
+    author attributions in that material or in the Appropriate Legal
+    Notices displayed by works containing it; or
+
+    c) Prohibiting misrepresentation of the origin of that material, or
+    requiring that modified versions of such material be marked in
+    reasonable ways as different from the original version; or
+
+    d) Limiting the use for publicity purposes of names of licensors or
+    authors of the material; or
+
+    e) Declining to grant rights under trademark law for use of some
+    trade names, trademarks, or service marks; or
+
+    f) Requiring indemnification of licensors and authors of that
+    material by anyone who conveys the material (or modified versions of
+    it) with contractual assumptions of liability to the recipient, for
+    any liability that these contractual assumptions directly impose on
+    those licensors and authors.
+
+  All other non-permissive additional terms are considered "further
+restrictions" within the meaning of section 10.  If the Program as you
+received it, or any part of it, contains a notice stating that it is
+governed by this License along with a term that is a further
+restriction, you may remove that term.  If a license document contains
+a further restriction but permits relicensing or conveying under this
+License, you may add to a covered work material governed by the terms
+of that license document, provided that the further restriction does
+not survive such relicensing or conveying.
+
+  If you add terms to a covered work in accord with this section, you
+must place, in the relevant source files, a statement of the
+additional terms that apply to those files, or a notice indicating
+where to find the applicable terms.
+
+  Additional terms, permissive or non-permissive, may be stated in the
+form of a separately written license, or stated as exceptions;
+the above requirements apply either way.
+
+  8. Termination.
+
+  You may not propagate or modify a covered work except as expressly
+provided under this License.  Any attempt otherwise to propagate or
+modify it is void, and will automatically terminate your rights under
+this License (including any patent licenses granted under the third
+paragraph of section 11).
+
+  However, if you cease all violation of this License, then your
+license from a particular copyright holder is reinstated (a)
+provisionally, unless and until the copyright holder explicitly and
+finally terminates your license, and (b) permanently, if the copyright
+holder fails to notify you of the violation by some reasonable means
+prior to 60 days after the cessation.
+
+  Moreover, your license from a particular copyright holder is
+reinstated permanently if the copyright holder notifies you of the
+violation by some reasonable means, this is the first time you have
+received notice of violation of this License (for any work) from that
+copyright holder, and you cure the violation prior to 30 days after
+your receipt of the notice.
+
+  Termination of your rights under this section does not terminate the
+licenses of parties who have received copies or rights from you under
+this License.  If your rights have been terminated and not permanently
+reinstated, you do not qualify to receive new licenses for the same
+material under section 10.
+
+  9. Acceptance Not Required for Having Copies.
+
+  You are not required to accept this License in order to receive or
+run a copy of the Program.  Ancillary propagation of a covered work
+occurring solely as a consequence of using peer-to-peer transmission
+to receive a copy likewise does not require acceptance.  However,
+nothing other than this License grants you permission to propagate or
+modify any covered work.  These actions infringe copyright if you do
+not accept this License.  Therefore, by modifying or propagating a
+covered work, you indicate your acceptance of this License to do so.
+
+  10. Automatic Licensing of Downstream Recipients.
+
+  Each time you convey a covered work, the recipient automatically
+receives a license from the original licensors, to run, modify and
+propagate that work, subject to this License.  You are not responsible
+for enforcing compliance by third parties with this License.
+
+  An "entity transaction" is a transaction transferring control of an
+organization, or substantially all assets of one, or subdividing an
+organization, or merging organizations.  If propagation of a covered
+work results from an entity transaction, each party to that
+transaction who receives a copy of the work also receives whatever
+licenses to the work the party's predecessor in interest had or could
+give under the previous paragraph, plus a right to possession of the
+Corresponding Source of the work from the predecessor in interest, if
+the predecessor has it or can get it with reasonable efforts.
+
+  You may not impose any further restrictions on the exercise of the
+rights granted or affirmed under this License.  For example, you may
+not impose a license fee, royalty, or other charge for exercise of
+rights granted under this License, and you may not initiate litigation
+(including a cross-claim or counterclaim in a lawsuit) alleging that
+any patent claim is infringed by making, using, selling, offering for
+sale, or importing the Program or any portion of it.
+
+  11. Patents.
+
+  A "contributor" is a copyright holder who authorizes use under this
+License of the Program or a work on which the Program is based.  The
+work thus licensed is called the contributor's "contributor version".
+
+  A contributor's "essential patent claims" are all patent claims
+owned or controlled by the contributor, whether already acquired or
+hereafter acquired, that would be infringed by some manner, permitted
+by this License, of making, using, or selling its contributor version,
+but do not include claims that would be infringed only as a
+consequence of further modification of the contributor version.  For
+purposes of this definition, "control" includes the right to grant
+patent sublicenses in a manner consistent with the requirements of
+this License.
+
+  Each contributor grants you a non-exclusive, worldwide, royalty-free
+patent license under the contributor's essential patent claims, to
+make, use, sell, offer for sale, import and otherwise run, modify and
+propagate the contents of its contributor version.
+
+  In the following three paragraphs, a "patent license" is any express
+agreement or commitment, however denominated, not to enforce a patent
+(such as an express permission to practice a patent or covenant not to
+sue for patent infringement).  To "grant" such a patent license to a
+party means to make such an agreement or commitment not to enforce a
+patent against the party.
+
+  If you convey a covered work, knowingly relying on a patent license,
+and the Corresponding Source of the work is not available for anyone
+to copy, free of charge and under the terms of this License, through a
+publicly available network server or other readily accessible means,
+then you must either (1) cause the Corresponding Source to be so
+available, or (2) arrange to deprive yourself of the benefit of the
+patent license for this particular work, or (3) arrange, in a manner
+consistent with the requirements of this License, to extend the patent
+license to downstream recipients.  "Knowingly relying" means you have
+actual knowledge that, but for the patent license, your conveying the
+covered work in a country, or your recipient's use of the covered work
+in a country, would infringe one or more identifiable patents in that
+country that you have reason to believe are valid.
+
+  If, pursuant to or in connection with a single transaction or
+arrangement, you convey, or propagate by procuring conveyance of, a
+covered work, and grant a patent license to some of the parties
+receiving the covered work authorizing them to use, propagate, modify
+or convey a specific copy of the covered work, then the patent license
+you grant is automatically extended to all recipients of the covered
+work and works based on it.
+
+  A patent license is "discriminatory" if it does not include within
+the scope of its coverage, prohibits the exercise of, or is
+conditioned on the non-exercise of one or more of the rights that are
+specifically granted under this License.  You may not convey a covered
+work if you are a party to an arrangement with a third party that is
+in the business of distributing software, under which you make payment
+to the third party based on the extent of your activity of conveying
+the work, and under which the third party grants, to any of the
+parties who would receive the covered work from you, a discriminatory
+patent license (a) in connection with copies of the covered work
+conveyed by you (or copies made from those copies), or (b) primarily
+for and in connection with specific products or compilations that
+contain the covered work, unless you entered into that arrangement,
+or that patent license was granted, prior to 28 March 2007.
+
+  Nothing in this License shall be construed as excluding or limiting
+any implied license or other defenses to infringement that may
+otherwise be available to you under applicable patent law.
+
+  12. No Surrender of Others' Freedom.
+
+  If conditions are imposed on you (whether by court order, agreement or
+otherwise) that contradict the conditions of this License, they do not
+excuse you from the conditions of this License.  If you cannot convey a
+covered work so as to satisfy simultaneously your obligations under this
+License and any other pertinent obligations, then as a consequence you may
+not convey it at all.  For example, if you agree to terms that obligate you
+to collect a royalty for further conveying from those to whom you convey
+the Program, the only way you could satisfy both those terms and this
+License would be to refrain entirely from conveying the Program.
+
+  13. Use with the GNU Affero General Public License.
+
+  Notwithstanding any other provision of this License, you have
+permission to link or combine any covered work with a work licensed
+under version 3 of the GNU Affero General Public License into a single
+combined work, and to convey the resulting work.  The terms of this
+License will continue to apply to the part which is the covered work,
+but the special requirements of the GNU Affero General Public License,
+section 13, concerning interaction through a network will apply to the
+combination as such.
+
+  14. Revised Versions of this License.
+
+  The Free Software Foundation may publish revised and/or new versions of
+the GNU General Public License from time to time.  Such new versions will
+be similar in spirit to the present version, but may differ in detail to
+address new problems or concerns.
+
+  Each version is given a distinguishing version number.  If the
+Program specifies that a certain numbered version of the GNU General
+Public License "or any later version" applies to it, you have the
+option of following the terms and conditions either of that numbered
+version or of any later version published by the Free Software
+Foundation.  If the Program does not specify a version number of the
+GNU General Public License, you may choose any version ever published
+by the Free Software Foundation.
+
+  If the Program specifies that a proxy can decide which future
+versions of the GNU General Public License can be used, that proxy's
+public statement of acceptance of a version permanently authorizes you
+to choose that version for the Program.
+
+  Later license versions may give you additional or different
+permissions.  However, no additional obligations are imposed on any
+author or copyright holder as a result of your choosing to follow a
+later version.
+
+  15. Disclaimer of Warranty.
+
+  THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY
+APPLICABLE LAW.  EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT
+HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY
+OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO,
+THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+PURPOSE.  THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM
+IS WITH YOU.  SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF
+ALL NECESSARY SERVICING, REPAIR OR CORRECTION.
+
+  16. Limitation of Liability.
+
+  IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
+WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS
+THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY
+GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE
+USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF
+DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD
+PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS),
+EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF
+SUCH DAMAGES.
+
+  17. Interpretation of Sections 15 and 16.
+
+  If the disclaimer of warranty and limitation of liability provided
+above cannot be given local legal effect according to their terms,
+reviewing courts shall apply local law that most closely approximates
+an absolute waiver of all civil liability in connection with the
+Program, unless a warranty or assumption of liability accompanies a
+copy of the Program in return for a fee.
+
+                     END OF TERMS AND CONDITIONS
+
+            How to Apply These Terms to Your New Programs
+
+  If you develop a new program, and you want it to be of the greatest
+possible use to the public, the best way to achieve this is to make it
+free software which everyone can redistribute and change under these terms.
+
+  To do so, attach the following notices to the program.  It is safest
+to attach them to the start of each source file to most effectively
+state the exclusion of warranty; and each file should have at least
+the "copyright" line and a pointer to where the full notice is found.
+
+    <one line to give the program's name and a brief idea of what it does.>
+    Copyright (C) <year>  <name of author>
+
+    This program is free software: you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation, either version 3 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License
+    along with this program.  If not, see <https://www.gnu.org/licenses/>.
+
+Also add information on how to contact you by electronic and paper mail.
+
+  If the program does terminal interaction, make it output a short
+notice like this when it starts in an interactive mode:
+
+    <program>  Copyright (C) <year>  <name of author>
+    This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'.
+    This is free software, and you are welcome to redistribute it
+    under certain conditions; type `show c' for details.
+
+The hypothetical commands `show w' and `show c' should show the appropriate
+parts of the General Public License.  Of course, your program's commands
+might be different; for a GUI interface, you would use an "about box".
+
+  You should also get your employer (if you work as a programmer) or school,
+if any, to sign a "copyright disclaimer" for the program, if necessary.
+For more information on this, and how to apply and follow the GNU GPL, see
+<https://www.gnu.org/licenses/>.
+
+  The GNU General Public License does not permit incorporating your program
+into proprietary programs.  If your program is a subroutine library, you
+may consider it more useful to permit linking proprietary applications with
+the library.  If this is what you want to do, use the GNU Lesser General
+Public License instead of this License.  But first, please read
+<https://www.gnu.org/licenses/why-not-lgpl.html>.

Abysz-LAB-Ext/README.md

@@ -0,0 +1,53 @@
+# Abysz deflicking & temporal coherence lab. 
+## Automatic1111 Extension. Beta 0.1.9
+
+![0 1 9](https://user-images.githubusercontent.com/112580728/228404036-5dbf11f1-51e3-4fb0-9c72-c3c2c0deb00e.png)
+|
+![ABYSZLAB09b](https://user-images.githubusercontent.com/112580728/226314389-ac838672-4af0-4d94-bde8-26fd83610a5f.png)
+
+## How DFI works:
+
+https://user-images.githubusercontent.com/112580728/226049549-e61bddb3-88ea-4953-893d-9993dd165180.mp4
+
+# Requirements
+
+OpenCV: ```pip install opencv-python```
+
+Imagemagick library: https://imagemagick.org/script/download.php
+
+## Basic guide:
+Differential frame interpolation analyzes the stability of the original video, and processes the generated video with that information. Example, if your original background is static, it will force the generated video to respect that, acting as a complex deflicker. It is an aggressive process, for which we need and will have a lot of control.
+
+Gui version 0.0.6 includes the following parameters.
+
+**Frame refresh frequency:** Every how many frames the interpolation is reduced. It allows to keep more information of the generated video, and avoid major ghosting.
+
+**Refresh Strength:** Opacity % of the interpolated information. 0 refreshes the entire frame, with no changes. Here you control how much change you allow overall.
+
+**DFI Strength:** Amount of information that tries to force. 4-6 recommended.
+
+**DFI Deghost:** A variable that generally reduces the areas affected by DFI. This can reduce ghosting without changing DFI strength.
+
+**Smooth:** Smoothes the interpolation. High values reduce the effectiveness of the process.
+
+**Source denoise:** Improves scanning in noisy sources.
+
+(DEFLICKERS PLAYGROUND ADDED)
+(FUSE AND VIDEO EXTRACT ADDED)
+
+# USE STRATEGIES:
+
+### Basic: 
+The simplest use is to find the balance between deflicking and deghosting. However, this is not efficient.
+
+## Multipass:
+The most efficient way to use this tool is to allow a certain amount of corruption and ghosting, in exchange for more stable video. Once we have that base, we must use a second step in Stable Diffusion, at low denoising (1-4). In most cases, this brings back much of the detail, but retains the stability we've gained.
+
+# Multibatch-controlnet: 
+The best, best way to use this tool is to use our "stabilized" video in img2img, and the original (REAL) video in controlnet HED. Then use a parallel batch to retrieve details. This considerably improves the multipass technique. Unfortunately, that function is not available in the controlnet gui as of this writing.
+
+# TODO
+Automatic1111 extension. Given my limited knowledge of programming, I had trouble getting my script to interact within A1111. I hope soon to solve details to integrate this tool.
+Also, there are many important utilities that are in development, waiting to be added soon, such as polar rendering (like "front/back", but more complex), gif viewer, source analysis, preprocessing, etc.
+
+

Abysz-LAB-Ext/scripts/Abysz_Lab.py

@@ -0,0 +1,1202 @@
+import gradio as gr
+import subprocess
+import os
+import imageio
+import numpy as np
+from gradio.outputs import Image
+from PIL import Image
+import sys
+import cv2
+import shutil
+import time
+import math
+
+from modules import shared
+from modules import scripts
+from modules import script_callbacks
+
+
+class Script(scripts.Script):
+    def title(self):
+        return "Abysz LAB"
+
+    def show(self, is_img2img):
+        return scripts.AlwaysVisible
+     
+    def ui(self, is_img2img):
+        return []
+        
+def main(ruta_entrada_1, ruta_entrada_2, ruta_salida, denoise_blur, dfi_strength, dfi_deghost, test_mode, inter_denoise, inter_denoise_size, inter_denoise_speed, fine_blur, frame_refresh_frequency, refresh_strength, smooth, frames_limit):
+            
+        maskD = os.path.join(os.getcwd(), 'extensions', 'Abysz-LAB-Ext', 'scripts', 'Run', 'MaskD')
+        maskS = os.path.join(os.getcwd(), 'extensions', 'Abysz-LAB-Ext', 'scripts', 'Run', 'MaskS')
+        #output = os.path.join(os.getcwd(), 'extensions', 'Abysz-LAB-Ext', 'scripts', 'Run', 'Output')
+        source = os.path.join(os.getcwd(), 'extensions', 'Abysz-LAB-Ext', 'scripts', 'Run', 'Source')
+        #gen = os.path.join(os.getcwd(), 'extensions', 'Abysz-LAB-Ext', 'scripts', 'Run', 'Gen')
+
+        # verificar si las carpetas existen y eliminarlas si es el caso
+        if os.path.exists(source): # verificar si existe la carpeta source
+            shutil.rmtree(source) # eliminar la carpeta source y su contenido
+        if os.path.exists(maskS): # verificar si existe la carpeta maskS
+            shutil.rmtree(maskS) # eliminar la carpeta maskS y su contenido
+        if os.path.exists(maskD): # verificar si existe la carpeta maskS
+            shutil.rmtree(maskD) # eliminar la carpeta maskS y su contenido
+                
+        os.makedirs(source, exist_ok=True)
+        os.makedirs(maskS, exist_ok=True)
+        os.makedirs(ruta_salida, exist_ok=True)
+        os.makedirs(maskD, exist_ok=True)
+        #os.makedirs(gen, exist_ok=True)
+        
+        
+        def copy_images(ruta_entrada_1, ruta_entrada_2, frames_limit=0):
+            # Copiar todas las imágenes de la carpeta ruta_entrada_1 a la carpeta Source
+            count = 0
+            
+            archivos = os.listdir(ruta_entrada_1)
+            archivos_ordenados = sorted(archivos)
+            
+            for i, file in enumerate(archivos_ordenados):
+                if file.endswith(".jpg") or file.endswith(".jpeg") or file.endswith(".png"):
+                    img = Image.open(os.path.join(ruta_entrada_1, file))
+                    rgb_img = img.convert('RGB')
+                    rgb_img.save(os.path.join("./extensions/Abysz-LAB-Ext/scripts/Run/Source", "{:04d}.jpeg".format(i+1)), "jpeg", quality=100)
+                    count += 1
+                    if frames_limit > 0 and count >= frames_limit:
+                        break
+                                     
+                        
+        # Llamar a la función copy_images para copiar las imágenes
+        copy_images(ruta_entrada_1,ruta_salida, frames_limit)
+        
+        def sresize(ruta_entrada_2):
+            gen_folder = ruta_entrada_2
+            
+            # Carpeta donde se encuentran las imágenes de FULL
+            full_folder = "./extensions/Abysz-LAB-Ext/scripts/Run/Source"
+            
+            # Obtener la primera imagen en la carpeta Gen
+            gen_images = os.listdir(gen_folder)
+            gen_image_path = os.path.join(gen_folder, gen_images[0])
+            gen_image = cv2.imread(gen_image_path)
+            gen_height, gen_width = gen_image.shape[:2]
+            gen_aspect_ratio = gen_width / gen_height
+            
+            # Recorrer todas las imágenes en la carpeta FULL
+            for image_name in sorted(os.listdir(full_folder)): 
+                image_path = os.path.join(full_folder, image_name)
+                image = cv2.imread(image_path)
+                height, width = image.shape[:2]
+                aspect_ratio = width / height
+            
+                if aspect_ratio != gen_aspect_ratio:
+                    if aspect_ratio > gen_aspect_ratio:
+                        # La imagen es más ancha que la imagen de Gen
+                        crop_width = int(height * gen_aspect_ratio)
+                        x = int((width - crop_width) / 2)
+                        image = image[:, x:x+crop_width]
+                    else:
+                        # La imagen es más alta que la imagen de Gen
+                        crop_height = int(width / gen_aspect_ratio)
+                        y = int((height - crop_height) / 2)
+                        image = image[y:y+crop_height, :]
+            
+                # Redimensionar la imagen de FULL a la resolución de la imagen de Gen
+                image = cv2.resize(image, (gen_width, gen_height))
+            
+                # Guardar la imagen redimensionada en la carpeta FULL
+                cv2.imwrite(os.path.join(full_folder, image_name), image)
+        
+        sresize(ruta_entrada_2)
+            
+        def s_g_rename(ruta_entrada_2):
+                                
+            gen_dir = ruta_entrada_2 # ruta de la carpeta "Source"
+            
+            # Obtener una lista de los nombres de archivo en la carpeta ruta_entrada_2
+            files2 = os.listdir(gen_dir)
+            files2 = sorted(files2) # ordenar alfabéticamente la lista
+            # Renombrar cada archivo
+            for i, file_name in enumerate(files2):
+                old_path = os.path.join(gen_dir, file_name) # ruta actual del archivo
+                new_file_name = f"{i+1:04d}rename" # nuevo nombre de archivo con formato %04d
+                new_path = os.path.join(gen_dir, new_file_name + os.path.splitext(file_name)[1]) # nueva ruta del archivo
+                try:
+                    os.rename(old_path, new_path)
+                except FileExistsError:
+                    print(f"El archivo {new_file_name} ya existe. Se omite su renombre.")
+                    
+            # Obtener una lista de los nombres de archivo en la carpeta ruta_entrada_2
+            files2 = os.listdir(gen_dir)
+            files2 = sorted(files2) # ordenar alfabéticamente la lista
+            # Renombrar cada archivo
+            for i, file_name in enumerate(files2):
+                old_path = os.path.join(gen_dir, file_name) # ruta actual del archivo
+                new_file_name = f"{i+1:04d}" # nuevo nombre de archivo con formato %04d
+                new_path = os.path.join(gen_dir, new_file_name + os.path.splitext(file_name)[1]) # nueva ruta del archivo
+                try:
+                    os.rename(old_path, new_path)
+                except FileExistsError:
+                    print(f"El archivo {new_file_name} ya existe. Se omite su renombre.")
+        
+        s_g_rename(ruta_entrada_2)
+        
+        # Obtener el primer archivo de la carpeta ruta_entrada_2
+        gen_files = os.listdir(ruta_entrada_2)
+        if gen_files:
+            first_gen_file = gen_files[0]
+
+            # Copiar el archivo a la carpeta "Output" y reemplazar si ya existe
+            #output_file = "Output" + first_gen_file
+            #shutil.copyfile(ruta_entrada_2 + first_gen_file, output_file)
+            output_file = os.path.join(ruta_salida, first_gen_file)
+            shutil.copyfile(os.path.join(ruta_entrada_2, first_gen_file), output_file)
+        #subprocess call
+        def denoise(denoise_blur):
+            if denoise_blur < 1: # Condición 1: strength debe ser mayor a 1
+                return
+                         
+            denoise_kernel = denoise_blur
+            # Obtener la lista de nombres de archivos en la carpeta source
+            files = os.listdir("./extensions/Abysz-LAB-Ext/scripts/Run/Source")
+            
+            # Crear una carpeta destino si no existe
+            #if not os.path.exists("dest"):
+            #   os.mkdir("dest")
+            
+            # Recorrer cada archivo en la carpeta source
+            for file in files:
+                # Leer la imagen con opencv
+                img = cv2.imread(os.path.join("./extensions/Abysz-LAB-Ext/scripts/Run/Source", file))
+            
+                # Aplicar el filtro de blur con un tamaño de kernel 5x5
+                dst = cv2.bilateralFilter(img, denoise_kernel, 31, 31)
+                
+                # Eliminar el archivo original
+                #os.remove(os.path.join("SourceDFI", file))
+            
+                # Guardar la imagen resultante en la carpeta destino con el mismo nombre
+                cv2.imwrite(os.path.join("./extensions/Abysz-LAB-Ext/scripts/Run/Source", file), dst)
+                
+        denoise(denoise_blur)    
+        
+        # Definir la carpeta donde están los archivos
+        carpeta = './extensions/Abysz-LAB-Ext/scripts/Run/Source'
+        
+        # Crear la carpeta MaskD si no existe
+        os.makedirs('./extensions/Abysz-LAB-Ext/scripts/Run/MaskD', exist_ok=True)
+        
+        # Inicializar contador
+        contador = 1
+        
+        umbral_size = dfi_strength
+        # Iterar a través de los archivos de imagen en la carpeta Source
+        for filename in sorted(os.listdir(carpeta)):
+            # Cargar la imagen actual y la siguiente en escala de grises
+            if contador > 1:
+                siguiente = cv2.imread(os.path.join(carpeta, filename), cv2.IMREAD_GRAYSCALE)
+                diff = cv2.absdiff(anterior, siguiente)
+        
+                # Aplicar un umbral y guardar la imagen resultante en la carpeta MaskD. Menos es más.
+                umbral = umbral_size
+                umbralizado = cv2.threshold(diff, umbral, 255, cv2.THRESH_BINARY_INV)[1] # Invertir los colores
+                cv2.imwrite(os.path.join('./extensions/Abysz-LAB-Ext/scripts/Run/MaskD', f'{contador-1:04d}.png'), umbralizado)
+        
+            anterior = cv2.imread(os.path.join(carpeta, filename), cv2.IMREAD_GRAYSCALE)
+            contador += 1
+            
+            #Actualmente, el tipo de umbralización es cv2.THRESH_BINARY_INV, que invierte los colores de la imagen umbralizada. 
+            #Puedes cambiarlo a otro tipo de umbralización, 
+            #como cv2.THRESH_BINARY, cv2.THRESH_TRUNC, cv2.THRESH_TOZERO o cv2.THRESH_TOZERO_INV.
+        
+        
+        # Obtener la lista de los nombres de los archivos en la carpeta MaskD
+        files = os.listdir("./extensions/Abysz-LAB-Ext/scripts/Run/MaskD")
+        # Definir la carpeta donde están los archivos
+        carpeta = "./extensions/Abysz-LAB-Ext/scripts/Run/MaskD"
+        blur_kernel = smooth
+        
+        # Iterar sobre cada archivo
+        for file in files:
+            if dfi_deghost == 0:
+                
+                continue
+            # Leer la imagen de la carpeta MaskD
+            #img = cv2.imread("MaskD" + file)
+            img = cv2.imread(os.path.join("./extensions/Abysz-LAB-Ext/scripts/Run/MaskD", file))
+            
+            # Invertir la imagen usando la función bitwise_not()
+            img_inv = cv2.bitwise_not(img)
+            
+            kernel_size = dfi_deghost
+            
+            # Dilatar la imagen usando la función dilate()
+            kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (kernel_size, kernel_size)) # Puedes cambiar el tamaño y la forma del kernel según tus preferencias
+            img_dil = cv2.dilate(img_inv, kernel)
+            
+            # Volver a invertir la imagen usando la función bitwise_not()
+            img_out = cv2.bitwise_not(img_dil)
+            
+            # Sobrescribir la imagen en la carpeta MaskD con el mismo nombre que el original
+            #cv2.imwrite("MaskD" + file, img_out)
+            #cv2.imwrite(os.path.join("MaskD", file, img_out))
+            filename = os.path.join("./extensions/Abysz-LAB-Ext/scripts/Run/MaskD", file)
+            cv2.imwrite(filename, img_out)
+    
+        # Iterar a través de los archivos de imagen en la carpeta MaskD
+        if smooth > 1:
+            for imagen in os.listdir(carpeta):
+                if imagen.endswith(".jpg") or imagen.endswith(".png") or imagen.endswith(".jpeg"):
+                    # Leer la imagen
+                    img = cv2.imread(os.path.join(carpeta, imagen))
+                    # Aplicar el filtro
+                    img = cv2.GaussianBlur(img, (blur_kernel,blur_kernel),0)
+                    # Guardar la imagen con el mismo nombre
+                    cv2.imwrite(os.path.join(carpeta, imagen), img)
+        
+        
+        # INICIO DEL BATCH Obtener el nombre del archivo en MaskD sin ninguna extensión
+        # Agregar una variable de contador de bucles
+        loop_count = 0
+        
+        # Agregar un bucle while para ejecutar el código en bucle infinito
+        while True:
+        
+            mask_files = sorted(os.listdir(maskD))
+            if not mask_files:
+                print(f"No frames left")
+                # Eliminar las carpetas Source, MaskS y MaskD si no hay más archivos para procesar
+                shutil.rmtree(maskD)
+                shutil.rmtree(maskS)
+                shutil.rmtree(source)
+                break
+            
+            extra_mod = fine_blur
+            
+            mask = mask_files[0]
+            maskname = os.path.splitext(mask)[0]
+            
+            maskp_path = os.path.join(maskD, mask) 
+
+            img = cv2.imread(maskp_path, cv2.IMREAD_GRAYSCALE) # leer la imagen en escala de grises
+            n_white_pix = np.sum(img == 255) # contar los píxeles que son iguales a 255 (blanco)
+            total_pix = img.size # obtener el número total de píxeles en la imagen
+            percentage = (n_white_pix / total_pix) * 100 # calcular el porcentaje de píxeles blancos
+            percentage = round(percentage, 1) # redondear el porcentaje a 1 decimal
+            
+            # calcular la variable extra
+            extra = 100 - percentage # restar el porcentaje a 100
+            extra = extra / 3 # dividir el resultado por 3
+            extra = math.ceil(extra) # redondear hacia arriba al entero más cercano
+            if extra % 2 == 0: # verificar si el número es par
+                extra = extra + 1 # sumarle 1 para hacerlo impar
+            
+            # Dynamic Blur
+            imgb = cv2.imread(maskp_path) # leer la imagen con opencv
+            img_blur = cv2.GaussianBlur(imgb, (extra,extra),0)
+
+            # guardar la imagen modificada con el mismo nombre y ruta
+            cv2.imwrite(maskp_path, img_blur)
+
+            # Obtener la ruta de la imagen en la subcarpeta de output que tiene el mismo nombre que la imagen en MaskD
+            output_files = [f for f in os.listdir(ruta_salida) if os.path.splitext(f)[0] == maskname]
+            if not output_files:
+                print(f"No se encontró en {ruta_salida} una imagen con el mismo nombre que {maskname}.")
+                exit(1)
+            
+            output_file = os.path.join(ruta_salida, output_files[0])
+            
+            # Aplicar el comando magick composite con las opciones deseadas
+            composite_command = f"magick composite -compose CopyOpacity {os.path.join(maskD, mask)} {output_file} {os.path.join(maskS, 'result.png')}"
+            os.system(composite_command)
+            
+            # Obtener el nombre del archivo en output sin ninguna extensión
+            name = os.path.splitext(os.path.basename(output_file))[0]
+            
+            # Renombrar el archivo result.png con el nombre del archivo en output y la extensión .png
+            os.rename(os.path.join(maskS, 'result.png'), os.path.join(maskS, f"{name}.png"))
+            
+            #Guardar el directorio actual en una variable
+            original_dir = os.getcwd()
+            
+            #Cambiar al directorio de la carpeta MaskS
+            os.chdir(maskS)
+            
+            #Iterar a través de los archivos de imagen en la carpeta MaskS
+            for imagen in sorted(os.listdir(".")):
+                # Obtener el nombre de la imagen sin la extensión
+                nombre, extension = os.path.splitext(imagen)
+                # Obtener solo el número de la imagen
+                numero = ''.join(filter(str.isdigit, nombre))
+                # Definir el nombre de la siguiente imagen
+                siguiente = f"{int(numero)+1:0{len(numero)}}{extension}"
+                # Renombrar la imagen
+                os.rename(imagen, siguiente)
+            
+            # Volver al directorio original
+            os.chdir(original_dir)
+            
+            # Establecer un valor predeterminado para disolución
+            if frame_refresh_frequency < 1:
+                dissolve = percentage
+            else:
+                dissolve = 100 if loop_count % frame_refresh_frequency != 0 else refresh_strength
+                    
+        
+            # Obtener el nombre del archivo en MaskS sin la extensión
+            maskS_files = [f for f in os.listdir(maskS) if os.path.isfile(os.path.join(maskS, f)) and f.endswith('.png')]
+            if maskS_files:
+                filename = os.path.splitext(maskS_files[0])[0]
+            else:
+                print(f"No se encontraron archivos de imagen en la carpeta '{maskS}'")
+                filename = ''[0]
+            
+            # Salir del bucle si no hay más imágenes que procesar
+            if not filename:
+                break
+            
+            # Obtener la extensión del archivo en Gen con el mismo nombre
+            gen_files = [f for f in os.listdir(ruta_entrada_2) if os.path.isfile(os.path.join(ruta_entrada_2, f)) and f.startswith(filename)]
+            if gen_files:
+                ext = os.path.splitext(gen_files[0])[1]
+            else:
+                print(f"No se encontró ningún archivo con el nombre '{filename}' en la carpeta '{ruta_entrada_2}'")
+                ext = ''
+                                
+            # Componer la imagen de MaskS y Gen con disolución (si está definido) y guardarla en la carpeta de salida
+            os.system(f"magick composite {'-dissolve ' + str(dissolve) + '%' if dissolve is not None else ''} {maskS}/{filename}.png {ruta_entrada_2}/{filename}{ext} {ruta_salida}/{filename}{ext}")
+            
+            denoise_loop = inter_denoise_speed
+            kernel1 = inter_denoise
+            kernel2 = inter_denoise_size
+            
+            # Demo plus bilateral
+            if loop_count % denoise_loop == 0:
+                # listar archivos en la carpeta de salida
+                archivos = os.listdir(ruta_salida)
+                # obtener el último archivo
+                ultimo_archivo = os.path.join(ruta_salida, archivos[-1])
+                # cargar imagen con opencv
+                imagen = cv2.imread(ultimo_archivo)
+                # aplicar filtro bilateral
+                imagen_filtrada = cv2.bilateralFilter(imagen, kernel1, kernel2, kernel2)
+                # sobreescribir el original
+                cv2.imwrite(ultimo_archivo, imagen_filtrada)
+            
+            # Obtener el nombre del archivo más bajo en la carpeta MaskD
+            maskd_files = [f for f in os.listdir(maskD) if os.path.isfile(os.path.join(maskD, f)) and f.startswith('')]
+            if maskd_files:
+                maskd_file = os.path.join(maskD, sorted(maskd_files)[0])
+                os.remove(maskd_file)
+            
+            # Obtener el nombre del archivo más bajo en la carpeta MaskS
+            masks_files = [f for f in os.listdir(maskS) if os.path.isfile(os.path.join(maskS, f)) and f.startswith('')]
+            if masks_files:
+                masks_file = os.path.join(maskS, sorted(masks_files)[0])
+                os.remove(masks_file)
+                                
+            # Aumentar el contador de bucles
+            loop_count += 1
+            
+def dyndef(ruta_entrada_3, ruta_salida_1, ddf_strength):
+    if ddf_strength <= 0: # Condición 1: strength debe ser mayor a 0
+        return
+    imgs = []
+    files = sorted(os.listdir(ruta_entrada_3))
+    
+    for file in files:
+        img = cv2.imread(os.path.join(ruta_entrada_3, file))
+        imgs.append(img)
+    
+    for idx in range(len(imgs)-1, 0, -1):
+        current_img = imgs[idx]
+        prev_img = imgs[idx-1]
+        alpha = ddf_strength
+        
+        current_img = cv2.addWeighted(current_img, alpha, prev_img, 1-alpha, 0)
+        imgs[idx] = current_img
+        
+        if not os.path.exists(ruta_salida_1):
+            os.makedirs(ruta_salida_1)
+            
+        output_path = os.path.join(ruta_salida_1, files[idx]) # Usa el mismo nombre que el original
+        cv2.imwrite(output_path, current_img)
+    
+    # Copia el primer archivo de los originales al finalizar el proceso
+    shutil.copy(os.path.join(ruta_entrada_3, files[0]), os.path.join(ruta_salida_1, files[0]))
+
+
+
+def overlay_images(image1_path, image2_path, over_strength):
+      
+    opacity = over_strength
+    
+    # Abrir las imágenes
+    image1 = Image.open(image1_path).convert('RGBA')
+    image2 = Image.open(image2_path).convert('RGBA')
+
+    # Alinear el tamaño de las imágenes
+    if image1.size != image2.size:
+        image2 = image2.resize(image1.size)
+
+    # Convertir las imágenes en matrices NumPy
+    np_image1 = np.array(image1).astype(np.float64) / 255.0
+    np_image2 = np.array(image2).astype(np.float64) / 255.0
+
+    # Aplicar el método de fusión "overlay" a las imágenes
+    def basic(target, blend, opacity):
+        return target * opacity + blend * (1-opacity)
+
+    def blender(func):
+        def blend(target, blend, opacity=1, *args):
+            res = func(target, blend, *args)
+            res = basic(res, blend, opacity)
+            return np.clip(res, 0, 1)
+        return blend
+
+    class Blend:
+        @classmethod
+        def method(cls, name):
+            return getattr(cls, name)
+
+        normal = basic
+
+        @staticmethod
+        @blender
+        def overlay(target, blend, *args):
+            return  (target>0.5) * (1-(2-2*target)*(1-blend)) +\
+                    (target<=0.5) * (2*target*blend)
+
+    blended_image = Blend.overlay(np_image1, np_image2, opacity)
+
+    # Convertir la matriz de vuelta a una imagen PIL
+    blended_image = Image.fromarray((blended_image * 255).astype(np.uint8), 'RGBA').convert('RGB')
+
+    # Guardar la imagen resultante
+    return blended_image
+    
+def overlay_images2(image1_path, image2_path, fuse_strength):
+      
+    opacity = fuse_strength
+    
+    try:
+        image1 = Image.open(image1_path).convert('RGBA')
+        image2 = Image.open(image2_path).convert('RGBA')
+    except:
+        print("No more frames to fuse.")
+        return
+
+    # Alinear el tamaño de las imágenes
+    if image1.size != image2.size:
+        image1 = image1.resize(image2.size)
+
+    # Convertir las imágenes en matrices NumPy
+    np_image1 = np.array(image1).astype(np.float64) / 255.0
+    np_image2 = np.array(image2).astype(np.float64) / 255.0
+
+    # Aplicar el método de fusión "overlay" a las imágenes
+    def basic(target, blend, opacity):
+        return target * opacity + blend * (1-opacity)
+
+    def blender(func):
+        def blend(target, blend, opacity=1, *args):
+            res = func(target, blend, *args)
+            res = basic(res, blend, opacity)
+            return np.clip(res, 0, 1)
+        return blend
+
+    class Blend:
+        @classmethod
+        def method(cls, name):
+            return getattr(cls, name)
+
+        normal = basic
+
+        @staticmethod
+        @blender
+        def overlay(target, blend, *args):
+            return  (target>0.5) * (1-(2-2*target)*(1-blend)) +\
+                    (target<=0.5) * (2*target*blend)
+
+    blended_image = Blend.overlay(np_image1, np_image2, opacity)
+
+    # Convertir la matriz de vuelta a una imagen PIL
+    blended_image = Image.fromarray((blended_image * 255).astype(np.uint8), 'RGBA').convert('RGB')
+
+    # Guardar la imagen resultante
+    return blended_image
+
+def overlay_run(ruta_entrada_3, ruta_salida_1, ddf_strength, over_strength):
+    if over_strength <= 0: # Condición 1: strength debe ser mayor a 0
+            return
+       
+    # Si ddf_strength y/o over_strength son mayores a 0, utilizar ruta_salida_1 en lugar de ruta_entrada_3
+    if ddf_strength > 0:
+        ruta_entrada_3 = ruta_salida_1    
+        
+    if not os.path.exists("overtemp"):
+            os.makedirs("overtemp")
+            
+    if not os.path.exists(ruta_salida_1):
+            os.makedirs(ruta_salida_1)
+            
+    gen_path = ruta_entrada_3
+    images = sorted(os.listdir(gen_path))
+    image1_path = os.path.join(gen_path, images[0])
+    image2_path = os.path.join(gen_path, images[1])
+    
+     
+    fused_image = overlay_images(image1_path, image2_path, over_strength)
+    fuseover_path = "overtemp"
+    filename = os.path.basename(image1_path)
+    fused_image.save(os.path.join(fuseover_path, filename))
+    
+    
+    # Obtener una lista de todos los archivos en la carpeta "Gen"
+    gen_files = sorted(os.listdir(ruta_entrada_3))
+    
+    for i in range(len(gen_files) - 1):
+        image1_path = os.path.join(ruta_entrada_3, gen_files[i])
+        image2_path = os.path.join(ruta_entrada_3, gen_files[i+1])
+        blended_image = overlay_images(image1_path, image2_path, over_strength)
+        blended_image.save(os.path.join("overtemp", gen_files[i+1]))
+    
+    
+    # Definimos la ruta de la carpeta "overtemp"
+    ruta_overtemp = "overtemp"
+    
+    # Movemos todos los archivos de la carpeta "overtemp" a la carpeta "ruta_salida"
+    for archivo in os.listdir(ruta_overtemp):
+        origen = os.path.join(ruta_overtemp, archivo)
+        destino = os.path.join(ruta_salida_1, archivo)
+        shutil.move(origen, destino)
+        
+    # Ajustar contraste y brillo para cada imagen en la carpeta de entrada
+    if over_strength >= 0.4:
+        for nombre_archivo in os.listdir(ruta_salida_1):
+            # Cargar imagen
+            ruta_archivo = os.path.join(ruta_salida_1, nombre_archivo)
+            img = cv2.imread(ruta_archivo)
+        
+            # Ajustar contraste y brillo
+            alpha = 1  # Factor de contraste (mayor que 1 para aumentar el contraste)
+            beta = 10  # Valor de brillo (entero positivo para aumentar el brillo)
+            img_contrast = cv2.convertScaleAbs(img, alpha=alpha, beta=beta)
+        
+            # Guardar imagen resultante en la carpeta de salida
+            ruta_salida = os.path.join(ruta_salida_1, nombre_archivo)
+            cv2.imwrite(ruta_salida, img_contrast)
+
+def over_fuse(ruta_entrada_4, ruta_entrada_5, ruta_salida_2, fuse_strength):
+    # Obtener una lista de todos los archivos en la carpeta "Gen"
+    gen_files = os.listdir(ruta_entrada_4)
+    
+    # Ordenar la lista de archivos alfabéticamente
+    gen_files.sort()
+    
+    # Obtener una lista de todos los archivos en la carpeta "Source"
+    source_files = os.listdir(ruta_entrada_5)
+    
+    # Ordenar la lista de archivos alfabéticamente
+    source_files.sort()
+    
+    if not os.path.exists(ruta_salida_2):
+            os.makedirs(ruta_salida_2)
+            
+    for i in range(len(gen_files)):
+        image1_path = os.path.join(ruta_entrada_4, gen_files[i])
+        image2_path = os.path.join(ruta_entrada_5, source_files[i])
+        blended_image = overlay_images2(image1_path, image2_path, fuse_strength)
+        try:
+            blended_image.save(os.path.join(ruta_salida_2, gen_files[i]))
+        except Exception as e:
+            print("Error al guardar la imagen:", str(e))
+            print("No more frames to fuse")
+            break
+
+
+def norm(ruta_entrada_3, ruta_salida_1, ddf_strength, over_strength, norm_strength):
+    if norm_strength <= 0: # Condición 1: Norm_strength debe ser mayor a 0
+        return
+    
+    # Si ddf_strength y/o over_strength son mayores a 0, utilizar ruta_salida_1 en lugar de ruta_entrada_3
+    if ddf_strength > 0 or over_strength > 0:
+        ruta_entrada_3 = ruta_salida_1
+    
+    # Crear la carpeta GenOverNorm si no existe
+    if not os.path.exists("normtemp"):
+        os.makedirs("normtemp")
+        
+    if not os.path.exists(ruta_salida_1):
+        os.makedirs(ruta_salida_1)
+            
+    # Obtener una lista de todas las imágenes en la carpeta FuseOver
+    img_list = os.listdir(ruta_entrada_3)
+    img_list.sort() # Ordenar la lista en orden ascendente
+        
+    # Iterar a través de las imágenes
+    for i in range(len(img_list)-1):
+        # Cargar las dos imágenes a fusionar
+        img1 = cv2.imread(os.path.join(ruta_entrada_3, img_list[i]))
+        img2 = cv2.imread(os.path.join(ruta_entrada_3, img_list[i+1]))
+    
+        # Calcular la luminosidad promedio de cada imagen
+        avg1 = np.mean(cv2.cvtColor(img1, cv2.COLOR_BGR2GRAY))
+        avg2 = np.mean(cv2.cvtColor(img2, cv2.COLOR_BGR2GRAY))
+    
+        # Calcular los pesos para cada imagen
+        weight1 = avg1 / (avg1 + avg2)
+        weight2 = avg2 / (avg1 + avg2)
+    
+        # Fusionar las imágenes utilizando los pesos
+        result = cv2.addWeighted(img1, weight1, img2, weight2, 0)
+                                
+        # Guardar la imagen resultante en la carpeta GenOverNorm con el mismo nombre que la imagen original
+        cv2.imwrite(os.path.join("normtemp", img_list[i+1]), result)
+    
+    # Copiar la primera imagen en la carpeta GenOverNorm para mantener la secuencia completa
+    img0 = cv2.imread(os.path.join(ruta_entrada_3, img_list[0]))
+    cv2.imwrite(os.path.join("normtemp", img_list[0]), img0)
+    
+    # Definimos la ruta de la carpeta "overtemp"
+    ruta_overtemp = "normtemp"
+    
+    # Movemos todos los archivos de la carpeta "overtemp" a la carpeta "ruta_salida"
+    for archivo in os.listdir(ruta_overtemp):
+        origen = os.path.join(ruta_overtemp, archivo)
+        destino = os.path.join(ruta_salida_1, archivo)
+        shutil.move(origen, destino)
+        
+def deflickers(ruta_entrada_3, ruta_salida_1, ddf_strength, over_strength, norm_strength):
+    dyndef(ruta_entrada_3, ruta_salida_1, ddf_strength)
+    overlay_run(ruta_entrada_3, ruta_salida_1, ddf_strength, over_strength)
+    norm(ruta_entrada_3, ruta_salida_1, ddf_strength, over_strength, norm_strength)
+    
+def extract_video(ruta_entrada_6, ruta_salida_3, fps_count):
+
+    # Ruta del archivo de video
+    filename = ruta_entrada_6
+
+    # Directorio donde se guardarán los frames extraídos
+    output_dir = ruta_salida_3
+
+    # Abrir el archivo de video
+    cap = cv2.VideoCapture(filename)
+
+    # Obtener los FPS originales del video
+    fps = cap.get(cv2.CAP_PROP_FPS)
+
+    # Si fps_count es 0, utilizar los FPS originales
+    if fps_count == 0:
+        fps_count = fps
+
+    # Calcular el tiempo entre cada frame a extraer en milisegundos
+    frame_time = int(round(1000 / fps_count))
+
+    # Crear el directorio de salida si no existe
+    if not os.path.exists(output_dir):
+        os.makedirs(output_dir)
+
+    # Inicializar el contador de frames
+    frame_count = 0
+
+    # Inicializar el tiempo del último frame extraído
+    last_frame_time = 0
+
+    # Iterar sobre los frames del video
+    while True:
+        # Leer el siguiente frame
+        ret, frame = cap.read()
+
+        # Si no se pudo leer un frame, salir del loop
+        if not ret:
+            break
+
+        # Calcular el tiempo actual del frame en milisegundos
+        current_frame_time = int(round(cap.get(cv2.CAP_PROP_POS_MSEC)))
+
+        # Si todavía no ha pasado suficiente tiempo desde el último frame extraído, saltar al siguiente frame
+        if current_frame_time - last_frame_time < frame_time:
+            continue
+
+        # Incrementar el contador de frames
+        frame_count += 1
+
+        # Construir el nombre del archivo de salida
+        output_filename = os.path.join(output_dir, 'frame_{:04d}.jpeg'.format(frame_count))
+
+        # Guardar el frame como una imagen
+        cv2.imwrite(output_filename, frame)
+
+        # Actualizar el tiempo del último frame extraído
+        last_frame_time = current_frame_time
+
+    # Cerrar el archivo de video
+    cap.release()
+
+    # Mostrar información sobre el proceso finalizado
+    print("Extracted {} frames.".format(frame_count))
+
+def test_dfi(ruta_entrada_1, ruta_entrada_2, denoise_blur, dfi_strength, dfi_deghost, test_mode, smooth):
+        
+        
+        maskD = os.path.join(os.getcwd(), 'extensions', 'Abysz-LAB-Ext', 'scripts', 'Run', 'MaskDT')
+        #maskS = os.path.join(os.getcwd(), 'extensions', 'Abysz-LAB-Ext', 'scripts', 'Run', 'MaskST')
+        #output = os.path.join(os.getcwd(), 'extensions', 'Abysz-lab', 'scripts', 'Run', 'Output')
+        source = os.path.join(os.getcwd(), 'extensions', 'Abysz-LAB-Ext', 'scripts', 'Run', 'SourceT')
+        #gen = os.path.join(os.getcwd(), 'extensions', 'Abysz-LAB-Ext', 'scripts', 'Run', 'GenT')
+        
+        # verificar si las carpetas existen y eliminarlas si es el caso
+        if os.path.exists(source): # verificar si existe la carpeta source
+            shutil.rmtree(source) # eliminar la carpeta source y su contenido
+        #if os.path.exists(maskS): # verificar si existe la carpeta maskS
+        #    shutil.rmtree(maskS) # eliminar la carpeta maskS y su contenido
+        if os.path.exists(maskD): # verificar si existe la carpeta maskS
+            shutil.rmtree(maskD) # eliminar la carpeta maskS y su contenido
+        #if os.path.exists(gen): # verificar si existe la carpeta maskS
+        #    shutil.rmtree(gen) # eliminar la carpeta maskS y su contenido
+        #if os.path.exists(output): # verificar si existe la carpeta maskS
+        #    shutil.rmtree(output) # eliminar la carpeta maskS y su contenido
+            
+    
+        os.makedirs(source, exist_ok=True)
+        #os.makedirs(maskS, exist_ok=True)
+        #os.makedirs(output, exist_ok=True)
+        os.makedirs(maskD, exist_ok=True)
+        #os.makedirs(gen, exist_ok=True)
+        
+        
+        def copy_images(ruta_entrada_1, ruta_entrada_2):
+            if test_mode == 0:
+                # Usar el primer formato
+                indices = [10, 11, 20, 21, 30, 31] # Los índices de las imágenes que quieres copiar
+            else:
+                test_frames = test_mode
+                # Usar el segundo formato
+                indices = list(range(test_frames)) # Los primeros 30 índices
+            # Copiar todas las imágenes de la carpeta ruta_entrada_1 a la carpeta Source
+            for i in indices:
+                file = os.listdir(ruta_entrada_1)[i] # Obtener el nombre del archivo en el índice i
+                if file.endswith(".jpg") or file.endswith(".jpeg") or file.endswith(".png"): # Verificar que sea una imagen
+                    img = Image.open(os.path.join(ruta_entrada_1, file)) # Abrir la imagen
+                    rgb_img = img.convert('RGB') # Convertir a RGB
+                    rgb_img.save(os.path.join("./extensions/Abysz-LAB-Ext/scripts/Run/SourceT", "{:04d}.jpeg".format(i+1)), "jpeg", quality=100) # Guardar la imagen en la carpeta destino
+                  
+        # Llamar a la función copy_images para copiar las imágenes
+        copy_images(ruta_entrada_1, ruta_entrada_2)
+        
+        # Carpeta donde se encuentran las imágenes de Gen
+        def sresize(ruta_entrada_2):
+            gen_folder = ruta_entrada_2
+            
+            # Carpeta donde se encuentran las imágenes de FULL
+            full_folder = "./extensions/Abysz-LAB-Ext/scripts/Run/SourceT"
+            
+            # Obtener la primera imagen en la carpeta Gen
+            gen_images = os.listdir(gen_folder)
+            gen_image_path = os.path.join(gen_folder, gen_images[0])
+            gen_image = cv2.imread(gen_image_path)
+            gen_height, gen_width = gen_image.shape[:2]
+            gen_aspect_ratio = gen_width / gen_height
+            
+            # Recorrer todas las imágenes en la carpeta FULL
+            for image_name in os.listdir(full_folder):
+                image_path = os.path.join(full_folder, image_name)
+                image = cv2.imread(image_path)
+                height, width = image.shape[:2]
+                aspect_ratio = width / height
+            
+                if aspect_ratio != gen_aspect_ratio:
+                    if aspect_ratio > gen_aspect_ratio:
+                        # La imagen es más ancha que la imagen de Gen
+                        crop_width = int(height * gen_aspect_ratio)
+                        x = int((width - crop_width) / 2)
+                        image = image[:, x:x+crop_width]
+                    else:
+                        # La imagen es más alta que la imagen de Gen
+                        crop_height = int(width / gen_aspect_ratio)
+                        y = int((height - crop_height) / 2)
+                        image = image[y:y+crop_height, :]
+            
+                # Redimensionar la imagen de FULL a la resolución de la imagen de Gen
+                image = cv2.resize(image, (gen_width, gen_height))
+            
+                # Guardar la imagen redimensionada en la carpeta FULL
+                cv2.imwrite(os.path.join(full_folder, image_name), image)
+        
+        sresize(ruta_entrada_2)
+            
+        def denoise(denoise_blur):
+            if denoise_blur < 1:
+                return
+                
+            denoise_kernel = denoise_blur
+            # Obtener la lista de nombres de archivos en la carpeta source
+            files = os.listdir("./extensions/Abysz-LAB-Ext/scripts/Run/SourceT")
+            
+            # Crear una carpeta destino si no existe
+            #if not os.path.exists("dest"):
+            #   os.mkdir("dest")
+            
+            # Recorrer cada archivo en la carpeta source
+            for file in files:
+                # Leer la imagen con opencv
+                img = cv2.imread(os.path.join("./extensions/Abysz-LAB-Ext/scripts/Run/SourceT", file))
+            
+                # Aplicar el filtro de blur con un tamaño de kernel 5x5
+                dst = cv2.bilateralFilter(img, denoise_kernel, 31, 31)
+                
+                # Eliminar el archivo original
+                #os.remove(os.path.join("SourceDFI", file))
+            
+                # Guardar la imagen resultante en la carpeta destino con el mismo nombre
+                cv2.imwrite(os.path.join("./extensions/Abysz-LAB-Ext/scripts/Run/SourceT", file), dst)
+                
+        denoise(denoise_blur)    
+                  
+            
+        # Definir la carpeta donde están los archivos
+        carpeta = './extensions/Abysz-LAB-Ext/scripts/Run/SourceT'
+        
+        # Crear la carpeta MaskD si no existe
+        os.makedirs('./extensions/Abysz-LAB-Ext/scripts/Run/MaskDT', exist_ok=True)
+        
+        # Inicializar número de imagen
+        numero = 1
+        
+        umbral_size = dfi_strength
+        # Iterar a través de los archivos de imagen en la carpeta Source
+        for filename in sorted(os.listdir(carpeta)):
+            if test_mode == 0:
+                # Cargar la imagen actual en escala de grises
+                actual = cv2.imread(os.path.join(carpeta, filename), cv2.IMREAD_GRAYSCALE)
+                
+                # Si el número de imagen es par, procesar la imagen actual y la anterior
+                if numero % 2 == 0:
+                    diff = cv2.absdiff(anterior, actual)
+            
+                    # Aplicar un umbral y guardar la imagen resultante en la carpeta MaskD con el mismo nombre que el original. Menos es más.
+                    umbral = umbral_size
+                    umbralizado = cv2.threshold(diff, umbral, 255, cv2.THRESH_BINARY_INV)[1] # Invertir los colores
+                    cv2.imwrite(os.path.join('./extensions/Abysz-LAB-Ext/scripts/Run/MaskDT', filename), umbralizado)
+                    
+                # Guardar la imagen actual como anterior para el siguiente ciclo
+                anterior = actual
+                
+                # Incrementar el número de imagen para alternar entre pares e impares
+                numero += 1
+        
+            else:
+                       
+                # Iterar a través de los archivos de imagen en la carpeta Source
+                for filename in sorted(os.listdir(carpeta)):
+                    # Cargar la imagen actual y la siguiente en escala de grises
+                    
+                    if numero > 1:
+                        siguiente = cv2.imread(os.path.join(carpeta, filename), cv2.IMREAD_GRAYSCALE)
+                        diff = cv2.absdiff(anterior, siguiente)
+                
+                        # Aplicar un umbral y guardar la imagen resultante en la carpeta MaskD. Menos es más.
+                        umbral = umbral_size
+                        umbralizado = cv2.threshold(diff, umbral, 255, cv2.THRESH_BINARY_INV)[1] # Invertir los colores
+                        cv2.imwrite(os.path.join('./extensions/Abysz-LAB-Ext/scripts/Run/MaskDT', filename), umbralizado)
+                
+                    anterior = cv2.imread(os.path.join(carpeta, filename), cv2.IMREAD_GRAYSCALE)
+                    numero += 1    
+        
+              
+        
+        # Obtener la lista de los nombres de los archivos en la carpeta MaskD
+        files = os.listdir("./extensions/Abysz-LAB-Ext/scripts/Run/MaskDT")
+        # Definir la carpeta donde están los archivos
+        carpeta = "./extensions/Abysz-LAB-Ext/scripts/Run/MaskDT"
+        blur_kernel = smooth
+        
+        # Iterar sobre cada archivo
+        for file in files:
+            if dfi_deghost == 0:
+                
+                continue
+            # Leer la imagen de la carpeta MaskD
+            #img = cv2.imread("MaskD" + file)
+            img = cv2.imread(os.path.join("./extensions/Abysz-LAB-Ext/scripts/Run/MaskDT", file))
+            
+            # Invertir la imagen usando la función bitwise_not()
+            img_inv = cv2.bitwise_not(img)
+            
+            kernel_size = dfi_deghost
+            
+            # Dilatar la imagen usando la función dilate()
+            kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (kernel_size, kernel_size)) # Puedes cambiar el tamaño y la forma del kernel según tus preferencias
+            img_dil = cv2.dilate(img_inv, kernel)
+            
+            # Volver a invertir la imagen usando la función bitwise_not()
+            img_out = cv2.bitwise_not(img_dil)
+            
+            # Sobrescribir la imagen en la carpeta MaskD con el mismo nombre que el original
+            #cv2.imwrite("MaskD" + file, img_out)
+            #cv2.imwrite(os.path.join("MaskD", file, img_out))
+            filename = os.path.join("./extensions/Abysz-LAB-Ext/scripts/Run/MaskDT", file)
+            cv2.imwrite(filename, img_out)
+    
+        # Iterar a través de los archivos de imagen en la carpeta MaskD
+        if smooth > 1:
+            for imagen in os.listdir(carpeta):
+                if imagen.endswith(".jpg") or imagen.endswith(".png") or imagen.endswith(".jpeg"):
+                    # Leer la imagen
+                    img = cv2.imread(os.path.join(carpeta, imagen))
+                    # Aplicar el filtro
+                    img = cv2.GaussianBlur(img, (blur_kernel,blur_kernel),0)
+                    # Guardar la imagen con el mismo nombre
+                    cv2.imwrite(os.path.join(carpeta, imagen), img)
+                    
+        if test_mode == 0:         
+            nombres = os.listdir("./extensions/Abysz-LAB-Ext/scripts/Run/MaskDT") # obtener los nombres de los archivos en la carpeta MaskDT
+            ancho = 0 # variable para guardar el ancho acumulado de las ventanas
+            for i, nombre in enumerate(nombres): # recorrer cada nombre de archivo
+                imagen = cv2.imread("./extensions/Abysz-LAB-Ext/scripts/Run/MaskDT/" + nombre) # leer la imagen correspondiente
+                h, w, c = imagen.shape # obtener el alto, ancho y canales de la imagen
+                aspect_ratio = w / h # calcular la relación de aspecto
+                cv2.namedWindow(nombre, cv2.WINDOW_NORMAL) # crear una ventana con el nombre del archivo
+                ancho_ventana = 630 # definir un ancho fijo para las ventanas
+                alto_ventana = int(ancho_ventana / aspect_ratio) # calcular el alto proporcional al ancho y a la relación de aspecto
+                cv2.resizeWindow(nombre, ancho_ventana, alto_ventana) # cambiar el tamaño de la ventana según las dimensiones calculadas 
+                cv2.moveWindow(nombre, ancho, 0) # mover la ventana a una posición horizontal según el ancho acumulado 
+                cv2.imshow(nombre, imagen) # mostrar la imagen en la ventana
+                cv2.setWindowProperty(nombre,cv2.WND_PROP_TOPMOST,1.0) # poner la ventana en primer plano con un valor double 
+                ancho += ancho_ventana + 10 # aumentar el ancho acumulado en 410 píxeles para la siguiente ventana 
+            cv2.waitKey(4000) # esperar a que se presione una tecla para cerrar todas las ventanas 
+            cv2.destroyAllWindows() # cerrar todas las ventanas abiertas por OpenCV 
+        
+        
+        else:
+            
+            # Directorio de entrada de imágenes
+            ruta_entrada = "./extensions/Abysz-LAB-Ext/scripts/Run/MaskDT"
+            
+            # Obtener el tamaño de la primera imagen en el directorio de entrada
+            img_path = os.path.join(ruta_entrada, os.listdir(ruta_entrada)[0])
+            img = cv2.imread(img_path)
+            img_size = (img.shape[1], img.shape[0])
+            
+            # Fps del video
+            fps = 10
+            
+            # Crear objeto VideoWriter
+            fourcc = cv2.VideoWriter_fourcc(*'mp4v')
+            video_salida = cv2.VideoWriter('output.mp4', fourcc, fps, img_size)
+            
+            # Crear ventana con nombre "video"
+            cv2.namedWindow("video")
+            
+            # Establecer la ventana en primer plano
+            cv2.setWindowProperty("video", cv2.WND_PROP_TOPMOST,1.0)
+
+            # Crear ventana de visualización
+            # Leer imágenes en el directorio y agregarlas al video de salida
+            for file in sorted(os.listdir(ruta_entrada)):
+                if file.endswith(".jpg") or file.endswith(".jpeg") or file.endswith(".png"): # Verificar que sea una imagen
+                    img = cv2.imread(os.path.join(ruta_entrada, file)) # Leer la imagen
+                    #img_resized = cv2.resize(img, img_size) # Redimensionar la imagen
+                    video_salida.write(img) # Agregar la imagen al video
+                    
+            # Liberar el objeto VideoWriter
+            video_salida.release()
+            
+            # Crear objeto VideoCapture para leer el archivo de video recién creado
+            video_capture = cv2.VideoCapture('output.mp4')
+            
+            # Crear ventana con nombre "video"
+            cv2.namedWindow("video")
+            
+            # Establecer la ventana en primer plano
+            cv2.setWindowProperty("video", cv2.WND_PROP_FULLSCREEN, cv2.WINDOW_NORMAL)
+            
+            # Mostrar el video en una ventana
+            while True:
+                ret, img = video_capture.read()
+                if ret:
+                    cv2.imshow('video', img)
+                    cv2.waitKey(int(1000/fps))
+                else:
+                    break
+            
+            # Liberar el objeto VideoCapture y cerrar la ventana de visualización
+            video_capture.release()
+            cv2.destroyAllWindows()
+
+def dfi_video(ruta_salida):
+    # Directorio de entrada de imágenes
+    ruta_entrada = ruta_salida
+    
+    # Obtener el tamaño de la primera imagen en el directorio de entrada
+    img_path = os.path.join(ruta_entrada, os.listdir(ruta_entrada)[0])
+    img = cv2.imread(img_path)
+    img_size = (img.shape[1], img.shape[0])
+    
+    # Fps del video
+    fps = 15
+    
+    # Crear objeto VideoWriter
+    fourcc = cv2.VideoWriter_fourcc(*'mp4v')
+    video_salida = cv2.VideoWriter('output.mp4', fourcc, fps, img_size)
+    
+    # Crear ventana con nombre "video"
+    cv2.namedWindow("video")
+    
+    # Establecer la ventana en primer plano
+    cv2.setWindowProperty("video", cv2.WND_PROP_TOPMOST,1.0)
+
+    # Crear ventana de visualización
+    # Leer imágenes en el directorio y agregarlas al video de salida
+    for file in sorted(os.listdir(ruta_entrada)):
+        if file.endswith(".jpg") or file.endswith(".jpeg") or file.endswith(".png"): # Verificar que sea una imagen
+            img = cv2.imread(os.path.join(ruta_entrada, file)) # Leer la imagen
+            #img_resized = cv2.resize(img, img_size) # Redimensionar la imagen
+            video_salida.write(img) # Agregar la imagen al video
+            
+    # Liberar el objeto VideoWriter
+    video_salida.release()
+    
+    # Crear objeto VideoCapture para leer el archivo de video recién creado
+    video_capture = cv2.VideoCapture('output.mp4')
+    
+    # Crear ventana con nombre "video"
+    cv2.namedWindow("video")
+    
+    # Establecer la ventana en primer plano
+    cv2.setWindowProperty("video", cv2.WND_PROP_FULLSCREEN, cv2.WINDOW_NORMAL)
+    
+    # Mostrar el video en una ventana
+    while True:
+        ret, img = video_capture.read()
+        if ret:
+            cv2.imshow('video', img)
+            cv2.waitKey(int(1000/fps))
+        else:
+            break
+    
+    # Liberar el objeto VideoCapture y cerrar la ventana de visualización
+    video_capture.release()
+    cv2.destroyAllWindows()
+    
+    
+def add_tab():
+    print('LAB')
+    with gr.Blocks(analytics_enabled=False) as demo:
+        with gr.Tabs():
+            with gr.Tab("Main"):
+                with gr.Row():
+                    with gr.Column():
+                        with gr.Column():
+                            gr.Markdown("# Abysz LAB 0.1.9 Temporal coherence tools")
+                            gr.Markdown("## DFI Render")
+                        with gr.Column():
+                            ruta_entrada_1 = gr.Textbox(label="Original/reference frames folder", placeholder="RAW frames, or generated ones. (Read the strategies in the guide)")
+                            ruta_entrada_2 = gr.Textbox(label="Generated frames folder", placeholder="The frames of AI generated video")
+                            ruta_salida = gr.Textbox(label="Output folder", placeholder="Remember that each generation overwrites previous frames in the same folder.")
+                        with gr.Accordion("Info", open=False):
+                            gr.Markdown("This process detects static areas between frames (white) and moving areas (black). Use preview map and you will understand this. Basically, it will force the white areas to stay the same on the next frame.")
+                            gr.Markdown("DFI Tolerance adjusts how stiff this process is. Higher = more rigidity + corruption. Lower = more flexible, less corruption, but allows more flick. ")
+                            gr.Markdown("As complement, you can clean the map, to reduce detail and noise, or fatten/expand the areas detected by DFI. It is better that you use preview many times to experience how it works.")
+                            gr.Markdown("### IMPORTANT: The general algorithm is optimized to maintain a balance between deflicking and corruption, so that it is easier to use StableDiffusion at low denoising to reconstruct lost detail while preserving the stability gained.")
+                        with gr.Row():
+                            denoise_blur = gr.Slider(minimum=0, maximum=30, value=0, step=1, label="Map Denoise")
+                            dfi_strength = gr.Slider(minimum=0.5, maximum=20, value=5, step=0.5, label="DFI Tolerance")
+                            dfi_deghost = gr.Slider(minimum=0, maximum=50, value=0, step=1, label="DFI Expand")
+                        with gr.Accordion("Info", open=False):
+                            gr.Markdown("Here you can preview examples of the motion map for those parameters. It is useful, for example, to adjust denoise if you see that it detects unnecessary graininess. Keep in mind that what you see represents movement between two frames.")
+                            gr.Markdown("A good balance point is to throttle DFI until you find just a few things in areas that should be static. If you force it to be TOO clean, it will mostly increase the overall corruption.")
+                        with gr.Row():
+                            dfi_test = gr.Button(value="Preview DFI Map")
+                            test_mode = gr.Slider(minimum=0, maximum=100, value=0, step=1, label="Preview amount. 0 = Quick shot")
+                        with gr.Accordion("Advanced", open=False):
+                            with gr.Accordion("Info", open=False):
+                                gr.Markdown("**Inter Denoise:** Reduces render pixelation generated by corruption. However, be careful. It's resource hungry, and might remove excess detail. Not recommended to change size or FPD, but to use Stable Diffusion to remove the pixelation later.")
+                                gr.Markdown("**Inter Blur:** Fine tunes the dynamic blur algorithm for DFI map. Lower = Stronger blur effects. Between 2-3 recommended.")
+                                gr.Markdown("**Corruption Refresh:** To reduce the distortion generated by the process, you can recover original information every X number of frames. Lower number = faster refresh.")
+                                gr.Markdown("**Corruption Preserve:** Here you decide how much corruption keep in each corruption refresh. Low values will recover more of the original frame, with its changes and flickering, in exchange for reducing corruption. You must find the balance that works best for your goal.")
+                                gr.Markdown("**Smooth:** This smoothes the edges of the interpolated areas. Low values are currently recommended until the algorithm is updated.")
+                            with gr.Row():
+                                inter_denoise = gr.Slider(minimum=1, maximum=25, value=9, step=1, label="Inter Denoise")
+                                inter_denoise_size = gr.Slider(minimum=1, maximum=25, value=9, step=2, label="Inter Denoise Size")
+                                inter_denoise_speed = gr.Slider(minimum=1, maximum=15, value=3, step=1, label="Inter Denoise FPD")
+                                fine_blur = gr.Slider(minimum=1, maximum=5, value=3, step=0.1, label="Inter Blur")
+                            gr.Markdown("### The new dynamic algorithm will handle these parameters. Activate them only for manual control.")
+                            with gr.Row():
+                                frame_refresh_frequency = gr.Slider(minimum=0, maximum=30, value=0, step=1, label="Corruption Refresh (Lower = Faster)")
+                                refresh_strength = gr.Slider(minimum=0, maximum=100, value=0, step=5, label="Corruption Preserve")
+                                smooth = gr.Slider(minimum=1, maximum=99, value=1, step=2, label="Smooth")
+                        with gr.Row():
+                            frames_limit = gr.Number(label="Frames to render. 0=ALL")
+                            run_button = gr.Button(value="Run DFI", variant="primary")
+                            output_placeholder = gr.Textbox(label="Status", placeholder="STAND BY...")
+                        video_dfi = gr.Button(value="Show output folder video")
+                    with gr.Column():
+                        with gr.Column():
+                            gr.Markdown("# |")
+                            gr.Markdown("## Deflickers Playground")
+                            with gr.Column():
+                                ruta_entrada_3 = gr.Textbox(label="Frames folder", placeholder="Frames to process")
+                                ruta_salida_1 = gr.Textbox(label="Output folder", placeholder="Processed frames")
+                                with gr.Accordion("Info", open=False):
+                                    gr.Markdown("I made this series of deflickers based on the standard that Vegas Pro includes. You can use them together or separately. Be careful when mixing them.")
+                                    gr.Markdown("**Blend:** Blends a percentage between frames. This can soften transitions and highlights. 50 is half of each frame. 80 or 20 are recommended values.")
+                                    gr.Markdown("**Overlay:** Use the overlay image blending mode. Note that it works particularly good at mid-high values, wich will modify the overall contrast. You will have to decide what works for you.")
+                                    gr.Markdown("**Normalize:** Calculates the average between frames to merge them. It may be more practical if you don't have a specific Blend deflicker value in mind.")
+                                ddf_strength = gr.Slider(minimum=0, maximum=1, value=0, step=0.01, label="BLEND (0=Off)")
+                                over_strength = gr.Slider(minimum=0, maximum=1, value=0, step=0.01, label="OVERLAY (0=Off)")
+                                norm_strength = gr.Slider(minimum=0, maximum=1, value=0, step=1, label="NORMALIZE (0=Off))")
+                                dfk_button = gr.Button(value="Deflickers")
+            with gr.Tab("LAB Tools"):
+                with gr.Column():
+                    gr.Markdown("## Style Fuse")
+                    with gr.Accordion("Info", open=False):
+                                    gr.Markdown("With this you can merge two sets of frames with overlay technique. For example, you can take a style video that is just lights and/or colors, and overlay it on top of another video.")
+                                    gr.Markdown("The resulting video will be useful for use in Img2Img Batch and that the AI render preserves these added color and lighting details, along with the details of the original video.")
+                    with gr.Row():
+                        ruta_entrada_4 = gr.Textbox(label="Style frames", placeholder="Style to fuse")
+                        ruta_entrada_5 = gr.Textbox(label="Video frames", placeholder="Frames to process")
+                    with gr.Row():
+                        ruta_salida_2 = gr.Textbox(label="Output folder", placeholder="Processed frames")   
+                        fuse_strength = gr.Slider(minimum=0.1, maximum=1, value=0.5, step=0.01, label="Fuse Strength")
+                        fuse_button = gr.Button(value="Fuse")
+                    gr.Markdown("## Video extract")
+                    with gr.Row():
+                        ruta_entrada_6 = gr.Textbox(label="Video path", placeholder="Remember to use same fps as generated video for DFI")
+                        ruta_salida_3 = gr.Textbox(label="Output folder", placeholder="Processed frames")
+                    with gr.Row():
+                        fps_count = gr.Number(label="Fps. 0=Original")
+                        vidextract_button = gr.Button(value="Extract")
+                    output_placeholder2 = gr.Textbox(label="Status", placeholder="STAND BY...") 
+            with gr.Tab("Guide"):
+                with gr.Column():
+                    gr.Markdown("# What DFI does?")
+                    with gr.Accordion("Info", open=False):
+                        gr.Markdown("DFI processing analyzes the motion of the original video, and attempts to force that information into the generated video. Demo on https://github.com/AbyszOne/Abysz-LAB-Ext")
+                        gr.Markdown("In short, this will reduce flicker in areas of the video that don't need to change, but SD does. For example, for a man smoking, leaning against a pole, it will detect that the pole is static, and will try to prevent it from changing as much as possible.")
+                        gr.Markdown("This is an aggressive process that requires a lot of control for each context. Read the recommended strategies.")
+                        gr.Markdown("Although Video to Video is the most efficient way, a DFI One Shot method is under experimental development as well.")
+                    gr.Markdown("# Usage strategies")
+                    with gr.Accordion("Info", open=False):
+                        gr.Markdown("If you get enough understanding of the tool, you can achieve a much more stable and clean enough rendering. However, this is quite demanding.")
+                        gr.Markdown("Instead, a much friendlier and faster way to use this tool is as an intermediate step. For this, you can allow a reasonable degree of corruption in exchange for more general stability. ")
+                        gr.Markdown("You can then clean up the corruption and recover details with a second step in Stable Diffusion at low denoising (0.2-0.4), using the same parameters and seed.")
+                        gr.Markdown("In this way, the final result will have the stability that we have gained, maintaining final detail. If you find a balanced workflow, you will get something at least much more coherent and stable than the raw AI render.")
+                        gr.Markdown("**OPTIONAL:** Although not ideal, you can use the same AI generated video as the source, instead of the RAW. The trick is to use DFI and denoise to wash out map details so that you reduce low/mid changes between frames. If you only need a soft deflick, it is a valid option.")
+        
+        dt_inputs=[ruta_entrada_1, ruta_entrada_2, denoise_blur, dfi_strength, dfi_deghost, test_mode, smooth]
+        run_inputs=[ruta_entrada_1, ruta_entrada_2, ruta_salida, denoise_blur, dfi_strength, dfi_deghost, test_mode, inter_denoise, inter_denoise_size, inter_denoise_speed, fine_blur, frame_refresh_frequency, refresh_strength, smooth, frames_limit]
+        dfk_inputs=[ruta_entrada_3, ruta_salida_1, ddf_strength, over_strength, norm_strength]
+        fuse_inputs=[ruta_entrada_4, ruta_entrada_5, ruta_salida_2, fuse_strength]
+        ve_inputs=[ruta_entrada_6, ruta_salida_3, fps_count]
+        
+        dfi_test.click(fn=test_dfi, inputs=dt_inputs, outputs=output_placeholder)
+        run_button.click(fn=main, inputs=run_inputs, outputs=output_placeholder)
+        video_dfi.click(fn=dfi_video, inputs=ruta_salida, outputs=output_placeholder)
+        dfk_button.click(fn=deflickers, inputs=dfk_inputs, outputs=output_placeholder)
+        fuse_button.click(fn=over_fuse, inputs=fuse_inputs, outputs=output_placeholder2)
+        vidextract_button.click(fn=extract_video, inputs=ve_inputs, outputs=output_placeholder2)
+    return [(demo, "Abysz LAB", "demo")]
+        
+script_callbacks.on_ui_tabs(add_tab)

SD-CN-Animation/.gitignore

@@ -0,0 +1,6 @@
+__pycache__/
+out/
+videos/
+FP_Res/
+result.mp4
+*.pth

SD-CN-Animation/FloweR/model.py

@@ -0,0 +1,191 @@
+import torch
+import torch.nn as nn
+import torch.functional as F
+
+# Define the model
+class FloweR(nn.Module):
+  def __init__(self, input_size = (384, 384), window_size = 4):
+    super(FloweR, self).__init__()
+
+    self.input_size = input_size
+    self.window_size = window_size
+
+    # 2 channels for optical flow
+    # 1 channel for occlusion mask
+    # 3 channels for next frame prediction 
+    self.out_channels = 6
+    
+
+    #INPUT: 384 x 384 x 4 * 3 
+
+    ### DOWNSCALE ###
+    self.conv_block_1 = nn.Sequential(
+      nn.Conv2d(3 * self.window_size, 128, kernel_size=3, stride=1, padding='same'),
+      nn.ReLU(),
+    ) # 384 x 384 x 128
+
+    self.conv_block_2 = nn.Sequential(
+      nn.AvgPool2d(2),
+      nn.Conv2d(128, 128, kernel_size=3, stride=1, padding='same'),
+      nn.ReLU(),
+    ) # 192 x 192 x 128
+
+    self.conv_block_3 = nn.Sequential(
+      nn.AvgPool2d(2),
+      nn.Conv2d(128, 128, kernel_size=3, stride=1, padding='same'),
+      nn.ReLU(),
+    ) # 96 x 96 x 128
+
+    self.conv_block_4 = nn.Sequential(
+      nn.AvgPool2d(2),
+      nn.Conv2d(128, 128, kernel_size=3, stride=1, padding='same'),
+      nn.ReLU(),
+    ) # 48 x 48 x 128
+
+    self.conv_block_5 = nn.Sequential(
+      nn.AvgPool2d(2),
+      nn.Conv2d(128, 128, kernel_size=3, stride=1, padding='same'),
+      nn.ReLU(),
+    ) # 24 x 24 x 128
+
+    self.conv_block_6 = nn.Sequential(
+      nn.AvgPool2d(2),
+      nn.Conv2d(128, 128, kernel_size=3, stride=1, padding='same'),
+      nn.ReLU(),
+    ) # 12 x 12 x 128
+
+    self.conv_block_7 = nn.Sequential(
+      nn.AvgPool2d(2),
+      nn.Conv2d(128, 128, kernel_size=3, stride=1, padding='same'),
+      nn.ReLU(),
+    ) # 6 x 6 x 128
+
+    self.conv_block_8 = nn.Sequential(
+      nn.AvgPool2d(2),
+      nn.Conv2d(128, 128, kernel_size=3, stride=1, padding='same'),
+      nn.ReLU(),
+    ) # 3 x 3 x 128 - 9 input tokens
+
+    ### Transformer part ###
+    # To be done
+
+    ### UPSCALE ###
+    self.conv_block_9 = nn.Sequential(
+      nn.Upsample(scale_factor=2),
+      nn.Conv2d(128, 128, kernel_size=3, stride=1, padding='same'),
+      nn.ReLU(),
+    ) # 6 x 6 x 128
+
+    self.conv_block_10 = nn.Sequential(
+      nn.Upsample(scale_factor=2),
+      nn.Conv2d(128, 128, kernel_size=3, stride=1, padding='same'),
+      nn.ReLU(),
+    ) # 12 x 12 x 128
+
+    self.conv_block_11 = nn.Sequential(
+      nn.Upsample(scale_factor=2),
+      nn.Conv2d(128, 128, kernel_size=3, stride=1, padding='same'),
+      nn.ReLU(),
+    ) # 24 x 24 x 128
+
+    self.conv_block_12 = nn.Sequential(
+      nn.Upsample(scale_factor=2),
+      nn.Conv2d(128, 128, kernel_size=3, stride=1, padding='same'),
+      nn.ReLU(),
+    ) # 48 x 48 x 128
+
+    self.conv_block_13 = nn.Sequential(
+      nn.Upsample(scale_factor=2),
+      nn.Conv2d(128, 128, kernel_size=3, stride=1, padding='same'),
+      nn.ReLU(),
+    ) # 96 x 96 x 128
+
+    self.conv_block_14 = nn.Sequential(
+      nn.Upsample(scale_factor=2),
+      nn.Conv2d(128, 128, kernel_size=3, stride=1, padding='same'),
+      nn.ReLU(),
+    ) # 192 x 192 x 128
+    
+    self.conv_block_15 = nn.Sequential(
+      nn.Upsample(scale_factor=2),
+      nn.Conv2d(128, 128, kernel_size=3, stride=1, padding='same'),
+      nn.ReLU(),
+    ) # 384 x 384 x 128
+
+    self.conv_block_16 = nn.Conv2d(128, self.out_channels, kernel_size=3, stride=1, padding='same')
+
+  def forward(self, input_frames):
+
+    if input_frames.size(1) != self.window_size: 
+      raise Exception(f'Shape of the input is not compatable. There should be exactly {self.window_size} frames in an input video.')
+
+    h, w = self.input_size
+    # batch, frames, height, width, colors
+    input_frames_permuted = input_frames.permute((0, 1, 4, 2, 3))
+    # batch, frames, colors, height, width
+
+    in_x = input_frames_permuted.reshape(-1, self.window_size * 3, self.input_size[0], self.input_size[1])
+
+    ### DOWNSCALE ###
+    block_1_out = self.conv_block_1(in_x)        # 384 x 384 x 128
+    block_2_out = self.conv_block_2(block_1_out) # 192 x 192 x 128
+    block_3_out = self.conv_block_3(block_2_out) # 96 x 96 x 128
+    block_4_out = self.conv_block_4(block_3_out) # 48 x 48 x 128
+    block_5_out = self.conv_block_5(block_4_out) # 24 x 24 x 128
+    block_6_out = self.conv_block_6(block_5_out) # 12 x 12 x 128
+    block_7_out = self.conv_block_7(block_6_out) # 6 x 6 x 128
+    block_8_out = self.conv_block_8(block_7_out) # 3 x 3 x 128
+
+    ### UPSCALE ###
+    block_9_out = block_7_out + self.conv_block_9(block_8_out)    # 6 x 6 x 128
+    block_10_out = block_6_out + self.conv_block_10(block_9_out)  # 12 x 12 x 128
+    block_11_out = block_5_out + self.conv_block_11(block_10_out) # 24 x 24 x 128
+    block_12_out = block_4_out + self.conv_block_12(block_11_out) # 48 x 48 x 128
+    block_13_out = block_3_out + self.conv_block_13(block_12_out) # 96 x 96 x 128
+    block_14_out = block_2_out + self.conv_block_14(block_13_out) # 192 x 192 x 128
+    block_15_out = block_1_out + self.conv_block_15(block_14_out) # 384 x 384 x 128
+
+    block_16_out = self.conv_block_16(block_15_out) # 384 x 384 x (2 + 1 + 3)
+    out = block_16_out.reshape(-1, self.out_channels, self.input_size[0], self.input_size[1])
+
+    ### for future model training ###
+    device = out.get_device()
+
+    pred_flow = out[:,:2,:,:] * 255  # (-255, 255)
+    pred_occl = (out[:,2:3,:,:] + 1) / 2 # [0, 1]
+    pred_next = out[:,3:6,:,:]
+
+    # Generate sampling grids
+
+    # Create grid to upsample input
+    '''    
+    d = torch.linspace(-1, 1, 8)
+    meshx, meshy = torch.meshgrid((d, d))
+    grid = torch.stack((meshy, meshx), 2)
+    grid = grid.unsqueeze(0) '''
+    
+    grid_y, grid_x = torch.meshgrid(torch.arange(0, h), torch.arange(0, w))
+    flow_grid = torch.stack((grid_x, grid_y), dim=0).float()
+    flow_grid = flow_grid.unsqueeze(0).to(device=device)
+    flow_grid = flow_grid + pred_flow
+
+    flow_grid[:, 0, :, :] = 2 * flow_grid[:, 0, :, :] / (w - 1) - 1
+    flow_grid[:, 1, :, :] = 2 * flow_grid[:, 1, :, :] / (h - 1) - 1
+    # batch, flow_chanels, height, width
+    flow_grid = flow_grid.permute(0, 2, 3, 1)
+    # batch, height, width, flow_chanels
+
+    previous_frame = input_frames_permuted[:, -1, :, :, :]
+    sampling_mode = "bilinear" if self.training else "nearest" 
+    warped_frame = torch.nn.functional.grid_sample(previous_frame, flow_grid, mode=sampling_mode, padding_mode="reflection", align_corners=False)
+    alpha_mask = torch.clip(pred_occl * 10, 0, 1) * 0.04
+    pred_next = torch.clip(pred_next, -1, 1)
+    warped_frame = torch.clip(warped_frame, -1, 1)
+    next_frame = pred_next * alpha_mask + warped_frame * (1 - alpha_mask)
+
+    res = torch.cat((pred_flow / 255, pred_occl * 2 - 1, next_frame), dim=1)
+
+    # batch, channels, height, width
+    res = res.permute((0, 2, 3, 1))
+    # batch, height, width, channels
+    return res

SD-CN-Animation/LICENSE

@@ -0,0 +1,22 @@
+License
+
+Copyright (c) 2023 Alexey Borsky
+
+The Software is subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
+
+This repository can only be used for personal/research/non-commercial purposes. 
+However, for commercial requests, please contact us directly at 
+borsky.alexey@gmail.com. This restriction applies only to the code itself, all
+derivative works made using this repository (i.e. images and video) can be 
+used for any purposes without restrictions. 

SD-CN-Animation/RAFT/LICENSE

@@ -0,0 +1,29 @@
+BSD 3-Clause License
+
+Copyright (c) 2020, princeton-vl
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+* Redistributions of source code must retain the above copyright notice, this
+  list of conditions and the following disclaimer.
+
+* Redistributions in binary form must reproduce the above copyright notice,
+  this list of conditions and the following disclaimer in the documentation
+  and/or other materials provided with the distribution.
+
+* Neither the name of the copyright holder nor the names of its
+  contributors may be used to endorse or promote products derived from
+  this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
+FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

SD-CN-Animation/RAFT/corr.py

@@ -0,0 +1,91 @@
+import torch
+import torch.nn.functional as F
+from RAFT.utils.utils import bilinear_sampler, coords_grid
+
+try:
+    import alt_cuda_corr
+except:
+    # alt_cuda_corr is not compiled
+    pass
+
+
+class CorrBlock:
+    def __init__(self, fmap1, fmap2, num_levels=4, radius=4):
+        self.num_levels = num_levels
+        self.radius = radius
+        self.corr_pyramid = []
+
+        # all pairs correlation
+        corr = CorrBlock.corr(fmap1, fmap2)
+
+        batch, h1, w1, dim, h2, w2 = corr.shape
+        corr = corr.reshape(batch*h1*w1, dim, h2, w2)
+        
+        self.corr_pyramid.append(corr)
+        for i in range(self.num_levels-1):
+            corr = F.avg_pool2d(corr, 2, stride=2)
+            self.corr_pyramid.append(corr)
+
+    def __call__(self, coords):
+        r = self.radius
+        coords = coords.permute(0, 2, 3, 1)
+        batch, h1, w1, _ = coords.shape
+
+        out_pyramid = []
+        for i in range(self.num_levels):
+            corr = self.corr_pyramid[i]
+            dx = torch.linspace(-r, r, 2*r+1, device=coords.device)
+            dy = torch.linspace(-r, r, 2*r+1, device=coords.device)
+            delta = torch.stack(torch.meshgrid(dy, dx), axis=-1)
+
+            centroid_lvl = coords.reshape(batch*h1*w1, 1, 1, 2) / 2**i
+            delta_lvl = delta.view(1, 2*r+1, 2*r+1, 2)
+            coords_lvl = centroid_lvl + delta_lvl
+
+            corr = bilinear_sampler(corr, coords_lvl)
+            corr = corr.view(batch, h1, w1, -1)
+            out_pyramid.append(corr)
+
+        out = torch.cat(out_pyramid, dim=-1)
+        return out.permute(0, 3, 1, 2).contiguous().float()
+
+    @staticmethod
+    def corr(fmap1, fmap2):
+        batch, dim, ht, wd = fmap1.shape
+        fmap1 = fmap1.view(batch, dim, ht*wd)
+        fmap2 = fmap2.view(batch, dim, ht*wd) 
+        
+        corr = torch.matmul(fmap1.transpose(1,2), fmap2)
+        corr = corr.view(batch, ht, wd, 1, ht, wd)
+        return corr  / torch.sqrt(torch.tensor(dim).float())
+
+
+class AlternateCorrBlock:
+    def __init__(self, fmap1, fmap2, num_levels=4, radius=4):
+        self.num_levels = num_levels
+        self.radius = radius
+
+        self.pyramid = [(fmap1, fmap2)]
+        for i in range(self.num_levels):
+            fmap1 = F.avg_pool2d(fmap1, 2, stride=2)
+            fmap2 = F.avg_pool2d(fmap2, 2, stride=2)
+            self.pyramid.append((fmap1, fmap2))
+
+    def __call__(self, coords):
+        coords = coords.permute(0, 2, 3, 1)
+        B, H, W, _ = coords.shape
+        dim = self.pyramid[0][0].shape[1]
+
+        corr_list = []
+        for i in range(self.num_levels):
+            r = self.radius
+            fmap1_i = self.pyramid[0][0].permute(0, 2, 3, 1).contiguous()
+            fmap2_i = self.pyramid[i][1].permute(0, 2, 3, 1).contiguous()
+
+            coords_i = (coords / 2**i).reshape(B, 1, H, W, 2).contiguous()
+            corr, = alt_cuda_corr.forward(fmap1_i, fmap2_i, coords_i, r)
+            corr_list.append(corr.squeeze(1))
+
+        corr = torch.stack(corr_list, dim=1)
+        corr = corr.reshape(B, -1, H, W)
+        return corr / torch.sqrt(torch.tensor(dim).float())

SD-CN-Animation/RAFT/extractor.py

@@ -0,0 +1,267 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+
+class ResidualBlock(nn.Module):
+    def __init__(self, in_planes, planes, norm_fn='group', stride=1):
+        super(ResidualBlock, self).__init__()
+  
+        self.conv1 = nn.Conv2d(in_planes, planes, kernel_size=3, padding=1, stride=stride)
+        self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, padding=1)
+        self.relu = nn.ReLU(inplace=True)
+
+        num_groups = planes // 8
+
+        if norm_fn == 'group':
+            self.norm1 = nn.GroupNorm(num_groups=num_groups, num_channels=planes)
+            self.norm2 = nn.GroupNorm(num_groups=num_groups, num_channels=planes)
+            if not stride == 1:
+                self.norm3 = nn.GroupNorm(num_groups=num_groups, num_channels=planes)
+        
+        elif norm_fn == 'batch':
+            self.norm1 = nn.BatchNorm2d(planes)
+            self.norm2 = nn.BatchNorm2d(planes)
+            if not stride == 1:
+                self.norm3 = nn.BatchNorm2d(planes)
+        
+        elif norm_fn == 'instance':
+            self.norm1 = nn.InstanceNorm2d(planes)
+            self.norm2 = nn.InstanceNorm2d(planes)
+            if not stride == 1:
+                self.norm3 = nn.InstanceNorm2d(planes)
+
+        elif norm_fn == 'none':
+            self.norm1 = nn.Sequential()
+            self.norm2 = nn.Sequential()
+            if not stride == 1:
+                self.norm3 = nn.Sequential()
+
+        if stride == 1:
+            self.downsample = None
+        
+        else:    
+            self.downsample = nn.Sequential(
+                nn.Conv2d(in_planes, planes, kernel_size=1, stride=stride), self.norm3)
+
+
+    def forward(self, x):
+        y = x
+        y = self.relu(self.norm1(self.conv1(y)))
+        y = self.relu(self.norm2(self.conv2(y)))
+
+        if self.downsample is not None:
+            x = self.downsample(x)
+
+        return self.relu(x+y)
+
+
+
+class BottleneckBlock(nn.Module):
+    def __init__(self, in_planes, planes, norm_fn='group', stride=1):
+        super(BottleneckBlock, self).__init__()
+  
+        self.conv1 = nn.Conv2d(in_planes, planes//4, kernel_size=1, padding=0)
+        self.conv2 = nn.Conv2d(planes//4, planes//4, kernel_size=3, padding=1, stride=stride)
+        self.conv3 = nn.Conv2d(planes//4, planes, kernel_size=1, padding=0)
+        self.relu = nn.ReLU(inplace=True)
+
+        num_groups = planes // 8
+
+        if norm_fn == 'group':
+            self.norm1 = nn.GroupNorm(num_groups=num_groups, num_channels=planes//4)
+            self.norm2 = nn.GroupNorm(num_groups=num_groups, num_channels=planes//4)
+            self.norm3 = nn.GroupNorm(num_groups=num_groups, num_channels=planes)
+            if not stride == 1:
+                self.norm4 = nn.GroupNorm(num_groups=num_groups, num_channels=planes)
+        
+        elif norm_fn == 'batch':
+            self.norm1 = nn.BatchNorm2d(planes//4)
+            self.norm2 = nn.BatchNorm2d(planes//4)
+            self.norm3 = nn.BatchNorm2d(planes)
+            if not stride == 1:
+                self.norm4 = nn.BatchNorm2d(planes)
+        
+        elif norm_fn == 'instance':
+            self.norm1 = nn.InstanceNorm2d(planes//4)
+            self.norm2 = nn.InstanceNorm2d(planes//4)
+            self.norm3 = nn.InstanceNorm2d(planes)
+            if not stride == 1:
+                self.norm4 = nn.InstanceNorm2d(planes)
+
+        elif norm_fn == 'none':
+            self.norm1 = nn.Sequential()
+            self.norm2 = nn.Sequential()
+            self.norm3 = nn.Sequential()
+            if not stride == 1:
+                self.norm4 = nn.Sequential()
+
+        if stride == 1:
+            self.downsample = None
+        
+        else:    
+            self.downsample = nn.Sequential(
+                nn.Conv2d(in_planes, planes, kernel_size=1, stride=stride), self.norm4)
+
+
+    def forward(self, x):
+        y = x
+        y = self.relu(self.norm1(self.conv1(y)))
+        y = self.relu(self.norm2(self.conv2(y)))
+        y = self.relu(self.norm3(self.conv3(y)))
+
+        if self.downsample is not None:
+            x = self.downsample(x)
+
+        return self.relu(x+y)
+
+class BasicEncoder(nn.Module):
+    def __init__(self, output_dim=128, norm_fn='batch', dropout=0.0):
+        super(BasicEncoder, self).__init__()
+        self.norm_fn = norm_fn
+
+        if self.norm_fn == 'group':
+            self.norm1 = nn.GroupNorm(num_groups=8, num_channels=64)
+            
+        elif self.norm_fn == 'batch':
+            self.norm1 = nn.BatchNorm2d(64)
+
+        elif self.norm_fn == 'instance':
+            self.norm1 = nn.InstanceNorm2d(64)
+
+        elif self.norm_fn == 'none':
+            self.norm1 = nn.Sequential()
+
+        self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3)
+        self.relu1 = nn.ReLU(inplace=True)
+
+        self.in_planes = 64
+        self.layer1 = self._make_layer(64,  stride=1)
+        self.layer2 = self._make_layer(96, stride=2)
+        self.layer3 = self._make_layer(128, stride=2)
+
+        # output convolution
+        self.conv2 = nn.Conv2d(128, output_dim, kernel_size=1)
+
+        self.dropout = None
+        if dropout > 0:
+            self.dropout = nn.Dropout2d(p=dropout)
+
+        for m in self.modules():
+            if isinstance(m, nn.Conv2d):
+                nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
+            elif isinstance(m, (nn.BatchNorm2d, nn.InstanceNorm2d, nn.GroupNorm)):
+                if m.weight is not None:
+                    nn.init.constant_(m.weight, 1)
+                if m.bias is not None:
+                    nn.init.constant_(m.bias, 0)
+
+    def _make_layer(self, dim, stride=1):
+        layer1 = ResidualBlock(self.in_planes, dim, self.norm_fn, stride=stride)
+        layer2 = ResidualBlock(dim, dim, self.norm_fn, stride=1)
+        layers = (layer1, layer2)
+        
+        self.in_planes = dim
+        return nn.Sequential(*layers)
+
+
+    def forward(self, x):
+
+        # if input is list, combine batch dimension
+        is_list = isinstance(x, tuple) or isinstance(x, list)
+        if is_list:
+            batch_dim = x[0].shape[0]
+            x = torch.cat(x, dim=0)
+
+        x = self.conv1(x)
+        x = self.norm1(x)
+        x = self.relu1(x)
+
+        x = self.layer1(x)
+        x = self.layer2(x)
+        x = self.layer3(x)
+
+        x = self.conv2(x)
+
+        if self.training and self.dropout is not None:
+            x = self.dropout(x)
+
+        if is_list:
+            x = torch.split(x, [batch_dim, batch_dim], dim=0)
+
+        return x
+
+
+class SmallEncoder(nn.Module):
+    def __init__(self, output_dim=128, norm_fn='batch', dropout=0.0):
+        super(SmallEncoder, self).__init__()
+        self.norm_fn = norm_fn
+
+        if self.norm_fn == 'group':
+            self.norm1 = nn.GroupNorm(num_groups=8, num_channels=32)
+            
+        elif self.norm_fn == 'batch':
+            self.norm1 = nn.BatchNorm2d(32)
+
+        elif self.norm_fn == 'instance':
+            self.norm1 = nn.InstanceNorm2d(32)
+
+        elif self.norm_fn == 'none':
+            self.norm1 = nn.Sequential()
+
+        self.conv1 = nn.Conv2d(3, 32, kernel_size=7, stride=2, padding=3)
+        self.relu1 = nn.ReLU(inplace=True)
+
+        self.in_planes = 32
+        self.layer1 = self._make_layer(32,  stride=1)
+        self.layer2 = self._make_layer(64, stride=2)
+        self.layer3 = self._make_layer(96, stride=2)
+
+        self.dropout = None
+        if dropout > 0:
+            self.dropout = nn.Dropout2d(p=dropout)
+        
+        self.conv2 = nn.Conv2d(96, output_dim, kernel_size=1)
+
+        for m in self.modules():
+            if isinstance(m, nn.Conv2d):
+                nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
+            elif isinstance(m, (nn.BatchNorm2d, nn.InstanceNorm2d, nn.GroupNorm)):
+                if m.weight is not None:
+                    nn.init.constant_(m.weight, 1)
+                if m.bias is not None:
+                    nn.init.constant_(m.bias, 0)
+
+    def _make_layer(self, dim, stride=1):
+        layer1 = BottleneckBlock(self.in_planes, dim, self.norm_fn, stride=stride)
+        layer2 = BottleneckBlock(dim, dim, self.norm_fn, stride=1)
+        layers = (layer1, layer2)
+    
+        self.in_planes = dim
+        return nn.Sequential(*layers)
+
+
+    def forward(self, x):
+
+        # if input is list, combine batch dimension
+        is_list = isinstance(x, tuple) or isinstance(x, list)
+        if is_list:
+            batch_dim = x[0].shape[0]
+            x = torch.cat(x, dim=0)
+
+        x = self.conv1(x)
+        x = self.norm1(x)
+        x = self.relu1(x)
+
+        x = self.layer1(x)
+        x = self.layer2(x)
+        x = self.layer3(x)
+        x = self.conv2(x)
+
+        if self.training and self.dropout is not None:
+            x = self.dropout(x)
+
+        if is_list:
+            x = torch.split(x, [batch_dim, batch_dim], dim=0)
+
+        return x

SD-CN-Animation/RAFT/raft.py

@@ -0,0 +1,144 @@
+import numpy as np
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+from RAFT.update import BasicUpdateBlock, SmallUpdateBlock
+from RAFT.extractor import BasicEncoder, SmallEncoder
+from RAFT.corr import CorrBlock, AlternateCorrBlock
+from RAFT.utils.utils import bilinear_sampler, coords_grid, upflow8
+
+try:
+    autocast = torch.cuda.amp.autocast
+except:
+    # dummy autocast for PyTorch < 1.6
+    class autocast:
+        def __init__(self, enabled):
+            pass
+        def __enter__(self):
+            pass
+        def __exit__(self, *args):
+            pass
+
+
+class RAFT(nn.Module):
+    def __init__(self, args):
+        super(RAFT, self).__init__()
+        self.args = args
+
+        if args.small:
+            self.hidden_dim = hdim = 96
+            self.context_dim = cdim = 64
+            args.corr_levels = 4
+            args.corr_radius = 3
+        
+        else:
+            self.hidden_dim = hdim = 128
+            self.context_dim = cdim = 128
+            args.corr_levels = 4
+            args.corr_radius = 4
+
+        if 'dropout' not in self.args:
+            self.args.dropout = 0
+
+        if 'alternate_corr' not in self.args:
+            self.args.alternate_corr = False
+
+        # feature network, context network, and update block
+        if args.small:
+            self.fnet = SmallEncoder(output_dim=128, norm_fn='instance', dropout=args.dropout)        
+            self.cnet = SmallEncoder(output_dim=hdim+cdim, norm_fn='none', dropout=args.dropout)
+            self.update_block = SmallUpdateBlock(self.args, hidden_dim=hdim)
+
+        else:
+            self.fnet = BasicEncoder(output_dim=256, norm_fn='instance', dropout=args.dropout)        
+            self.cnet = BasicEncoder(output_dim=hdim+cdim, norm_fn='batch', dropout=args.dropout)
+            self.update_block = BasicUpdateBlock(self.args, hidden_dim=hdim)
+
+    def freeze_bn(self):
+        for m in self.modules():
+            if isinstance(m, nn.BatchNorm2d):
+                m.eval()
+
+    def initialize_flow(self, img):
+        """ Flow is represented as difference between two coordinate grids flow = coords1 - coords0"""
+        N, C, H, W = img.shape
+        coords0 = coords_grid(N, H//8, W//8, device=img.device)
+        coords1 = coords_grid(N, H//8, W//8, device=img.device)
+
+        # optical flow computed as difference: flow = coords1 - coords0
+        return coords0, coords1
+
+    def upsample_flow(self, flow, mask):
+        """ Upsample flow field [H/8, W/8, 2] -> [H, W, 2] using convex combination """
+        N, _, H, W = flow.shape
+        mask = mask.view(N, 1, 9, 8, 8, H, W)
+        mask = torch.softmax(mask, dim=2)
+
+        up_flow = F.unfold(8 * flow, [3,3], padding=1)
+        up_flow = up_flow.view(N, 2, 9, 1, 1, H, W)
+
+        up_flow = torch.sum(mask * up_flow, dim=2)
+        up_flow = up_flow.permute(0, 1, 4, 2, 5, 3)
+        return up_flow.reshape(N, 2, 8*H, 8*W)
+
+
+    def forward(self, image1, image2, iters=12, flow_init=None, upsample=True, test_mode=False):
+        """ Estimate optical flow between pair of frames """
+
+        image1 = 2 * (image1 / 255.0) - 1.0
+        image2 = 2 * (image2 / 255.0) - 1.0
+
+        image1 = image1.contiguous()
+        image2 = image2.contiguous()
+
+        hdim = self.hidden_dim
+        cdim = self.context_dim
+
+        # run the feature network
+        with autocast(enabled=self.args.mixed_precision):
+            fmap1, fmap2 = self.fnet([image1, image2])        
+        
+        fmap1 = fmap1.float()
+        fmap2 = fmap2.float()
+        if self.args.alternate_corr:
+            corr_fn = AlternateCorrBlock(fmap1, fmap2, radius=self.args.corr_radius)
+        else:
+            corr_fn = CorrBlock(fmap1, fmap2, radius=self.args.corr_radius)
+
+        # run the context network
+        with autocast(enabled=self.args.mixed_precision):
+            cnet = self.cnet(image1)
+            net, inp = torch.split(cnet, [hdim, cdim], dim=1)
+            net = torch.tanh(net)
+            inp = torch.relu(inp)
+
+        coords0, coords1 = self.initialize_flow(image1)
+
+        if flow_init is not None:
+            coords1 = coords1 + flow_init
+
+        flow_predictions = []
+        for itr in range(iters):
+            coords1 = coords1.detach()
+            corr = corr_fn(coords1) # index correlation volume
+
+            flow = coords1 - coords0
+            with autocast(enabled=self.args.mixed_precision):
+                net, up_mask, delta_flow = self.update_block(net, inp, corr, flow)
+
+            # F(t+1) = F(t) + \Delta(t)
+            coords1 = coords1 + delta_flow
+
+            # upsample predictions
+            if up_mask is None:
+                flow_up = upflow8(coords1 - coords0)
+            else:
+                flow_up = self.upsample_flow(coords1 - coords0, up_mask)
+            
+            flow_predictions.append(flow_up)
+
+        if test_mode:
+            return coords1 - coords0, flow_up
+            
+        return flow_predictions

SD-CN-Animation/RAFT/update.py

@@ -0,0 +1,139 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+
+class FlowHead(nn.Module):
+    def __init__(self, input_dim=128, hidden_dim=256):
+        super(FlowHead, self).__init__()
+        self.conv1 = nn.Conv2d(input_dim, hidden_dim, 3, padding=1)
+        self.conv2 = nn.Conv2d(hidden_dim, 2, 3, padding=1)
+        self.relu = nn.ReLU(inplace=True)
+
+    def forward(self, x):
+        return self.conv2(self.relu(self.conv1(x)))
+
+class ConvGRU(nn.Module):
+    def __init__(self, hidden_dim=128, input_dim=192+128):
+        super(ConvGRU, self).__init__()
+        self.convz = nn.Conv2d(hidden_dim+input_dim, hidden_dim, 3, padding=1)
+        self.convr = nn.Conv2d(hidden_dim+input_dim, hidden_dim, 3, padding=1)
+        self.convq = nn.Conv2d(hidden_dim+input_dim, hidden_dim, 3, padding=1)
+
+    def forward(self, h, x):
+        hx = torch.cat([h, x], dim=1)
+
+        z = torch.sigmoid(self.convz(hx))
+        r = torch.sigmoid(self.convr(hx))
+        q = torch.tanh(self.convq(torch.cat([r*h, x], dim=1)))
+
+        h = (1-z) * h + z * q
+        return h
+
+class SepConvGRU(nn.Module):
+    def __init__(self, hidden_dim=128, input_dim=192+128):
+        super(SepConvGRU, self).__init__()
+        self.convz1 = nn.Conv2d(hidden_dim+input_dim, hidden_dim, (1,5), padding=(0,2))
+        self.convr1 = nn.Conv2d(hidden_dim+input_dim, hidden_dim, (1,5), padding=(0,2))
+        self.convq1 = nn.Conv2d(hidden_dim+input_dim, hidden_dim, (1,5), padding=(0,2))
+
+        self.convz2 = nn.Conv2d(hidden_dim+input_dim, hidden_dim, (5,1), padding=(2,0))
+        self.convr2 = nn.Conv2d(hidden_dim+input_dim, hidden_dim, (5,1), padding=(2,0))
+        self.convq2 = nn.Conv2d(hidden_dim+input_dim, hidden_dim, (5,1), padding=(2,0))
+
+
+    def forward(self, h, x):
+        # horizontal
+        hx = torch.cat([h, x], dim=1)
+        z = torch.sigmoid(self.convz1(hx))
+        r = torch.sigmoid(self.convr1(hx))
+        q = torch.tanh(self.convq1(torch.cat([r*h, x], dim=1)))        
+        h = (1-z) * h + z * q
+
+        # vertical
+        hx = torch.cat([h, x], dim=1)
+        z = torch.sigmoid(self.convz2(hx))
+        r = torch.sigmoid(self.convr2(hx))
+        q = torch.tanh(self.convq2(torch.cat([r*h, x], dim=1)))       
+        h = (1-z) * h + z * q
+
+        return h
+
+class SmallMotionEncoder(nn.Module):
+    def __init__(self, args):
+        super(SmallMotionEncoder, self).__init__()
+        cor_planes = args.corr_levels * (2*args.corr_radius + 1)**2
+        self.convc1 = nn.Conv2d(cor_planes, 96, 1, padding=0)
+        self.convf1 = nn.Conv2d(2, 64, 7, padding=3)
+        self.convf2 = nn.Conv2d(64, 32, 3, padding=1)
+        self.conv = nn.Conv2d(128, 80, 3, padding=1)
+
+    def forward(self, flow, corr):
+        cor = F.relu(self.convc1(corr))
+        flo = F.relu(self.convf1(flow))
+        flo = F.relu(self.convf2(flo))
+        cor_flo = torch.cat([cor, flo], dim=1)
+        out = F.relu(self.conv(cor_flo))
+        return torch.cat([out, flow], dim=1)
+
+class BasicMotionEncoder(nn.Module):
+    def __init__(self, args):
+        super(BasicMotionEncoder, self).__init__()
+        cor_planes = args.corr_levels * (2*args.corr_radius + 1)**2
+        self.convc1 = nn.Conv2d(cor_planes, 256, 1, padding=0)
+        self.convc2 = nn.Conv2d(256, 192, 3, padding=1)
+        self.convf1 = nn.Conv2d(2, 128, 7, padding=3)
+        self.convf2 = nn.Conv2d(128, 64, 3, padding=1)
+        self.conv = nn.Conv2d(64+192, 128-2, 3, padding=1)
+
+    def forward(self, flow, corr):
+        cor = F.relu(self.convc1(corr))
+        cor = F.relu(self.convc2(cor))
+        flo = F.relu(self.convf1(flow))
+        flo = F.relu(self.convf2(flo))
+
+        cor_flo = torch.cat([cor, flo], dim=1)
+        out = F.relu(self.conv(cor_flo))
+        return torch.cat([out, flow], dim=1)
+
+class SmallUpdateBlock(nn.Module):
+    def __init__(self, args, hidden_dim=96):
+        super(SmallUpdateBlock, self).__init__()
+        self.encoder = SmallMotionEncoder(args)
+        self.gru = ConvGRU(hidden_dim=hidden_dim, input_dim=82+64)
+        self.flow_head = FlowHead(hidden_dim, hidden_dim=128)
+
+    def forward(self, net, inp, corr, flow):
+        motion_features = self.encoder(flow, corr)
+        inp = torch.cat([inp, motion_features], dim=1)
+        net = self.gru(net, inp)
+        delta_flow = self.flow_head(net)
+
+        return net, None, delta_flow
+
+class BasicUpdateBlock(nn.Module):
+    def __init__(self, args, hidden_dim=128, input_dim=128):
+        super(BasicUpdateBlock, self).__init__()
+        self.args = args
+        self.encoder = BasicMotionEncoder(args)
+        self.gru = SepConvGRU(hidden_dim=hidden_dim, input_dim=128+hidden_dim)
+        self.flow_head = FlowHead(hidden_dim, hidden_dim=256)
+
+        self.mask = nn.Sequential(
+            nn.Conv2d(128, 256, 3, padding=1),
+            nn.ReLU(inplace=True),
+            nn.Conv2d(256, 64*9, 1, padding=0))
+
+    def forward(self, net, inp, corr, flow, upsample=True):
+        motion_features = self.encoder(flow, corr)
+        inp = torch.cat([inp, motion_features], dim=1)
+
+        net = self.gru(net, inp)
+        delta_flow = self.flow_head(net)
+
+        # scale mask to balence gradients
+        mask = .25 * self.mask(net)
+        return net, mask, delta_flow
+
+
+

SD-CN-Animation/RAFT/utils/augmentor.py

@@ -0,0 +1,246 @@
+import numpy as np
+import random
+import math
+from PIL import Image
+
+import cv2
+cv2.setNumThreads(0)
+cv2.ocl.setUseOpenCL(False)
+
+import torch
+from torchvision.transforms import ColorJitter
+import torch.nn.functional as F
+
+
+class FlowAugmentor:
+    def __init__(self, crop_size, min_scale=-0.2, max_scale=0.5, do_flip=True):
+        
+        # spatial augmentation params
+        self.crop_size = crop_size
+        self.min_scale = min_scale
+        self.max_scale = max_scale
+        self.spatial_aug_prob = 0.8
+        self.stretch_prob = 0.8
+        self.max_stretch = 0.2
+
+        # flip augmentation params
+        self.do_flip = do_flip
+        self.h_flip_prob = 0.5
+        self.v_flip_prob = 0.1
+
+        # photometric augmentation params
+        self.photo_aug = ColorJitter(brightness=0.4, contrast=0.4, saturation=0.4, hue=0.5/3.14)
+        self.asymmetric_color_aug_prob = 0.2
+        self.eraser_aug_prob = 0.5
+
+    def color_transform(self, img1, img2):
+        """ Photometric augmentation """
+
+        # asymmetric
+        if np.random.rand() < self.asymmetric_color_aug_prob:
+            img1 = np.array(self.photo_aug(Image.fromarray(img1)), dtype=np.uint8)
+            img2 = np.array(self.photo_aug(Image.fromarray(img2)), dtype=np.uint8)
+
+        # symmetric
+        else:
+            image_stack = np.concatenate([img1, img2], axis=0)
+            image_stack = np.array(self.photo_aug(Image.fromarray(image_stack)), dtype=np.uint8)
+            img1, img2 = np.split(image_stack, 2, axis=0)
+
+        return img1, img2
+
+    def eraser_transform(self, img1, img2, bounds=[50, 100]):
+        """ Occlusion augmentation """
+
+        ht, wd = img1.shape[:2]
+        if np.random.rand() < self.eraser_aug_prob:
+            mean_color = np.mean(img2.reshape(-1, 3), axis=0)
+            for _ in range(np.random.randint(1, 3)):
+                x0 = np.random.randint(0, wd)
+                y0 = np.random.randint(0, ht)
+                dx = np.random.randint(bounds[0], bounds[1])
+                dy = np.random.randint(bounds[0], bounds[1])
+                img2[y0:y0+dy, x0:x0+dx, :] = mean_color
+
+        return img1, img2
+
+    def spatial_transform(self, img1, img2, flow):
+        # randomly sample scale
+        ht, wd = img1.shape[:2]
+        min_scale = np.maximum(
+            (self.crop_size[0] + 8) / float(ht), 
+            (self.crop_size[1] + 8) / float(wd))
+
+        scale = 2 ** np.random.uniform(self.min_scale, self.max_scale)
+        scale_x = scale
+        scale_y = scale
+        if np.random.rand() < self.stretch_prob:
+            scale_x *= 2 ** np.random.uniform(-self.max_stretch, self.max_stretch)
+            scale_y *= 2 ** np.random.uniform(-self.max_stretch, self.max_stretch)
+        
+        scale_x = np.clip(scale_x, min_scale, None)
+        scale_y = np.clip(scale_y, min_scale, None)
+
+        if np.random.rand() < self.spatial_aug_prob:
+            # rescale the images
+            img1 = cv2.resize(img1, None, fx=scale_x, fy=scale_y, interpolation=cv2.INTER_LINEAR)
+            img2 = cv2.resize(img2, None, fx=scale_x, fy=scale_y, interpolation=cv2.INTER_LINEAR)
+            flow = cv2.resize(flow, None, fx=scale_x, fy=scale_y, interpolation=cv2.INTER_LINEAR)
+            flow = flow * [scale_x, scale_y]
+
+        if self.do_flip:
+            if np.random.rand() < self.h_flip_prob: # h-flip
+                img1 = img1[:, ::-1]
+                img2 = img2[:, ::-1]
+                flow = flow[:, ::-1] * [-1.0, 1.0]
+
+            if np.random.rand() < self.v_flip_prob: # v-flip
+                img1 = img1[::-1, :]
+                img2 = img2[::-1, :]
+                flow = flow[::-1, :] * [1.0, -1.0]
+
+        y0 = np.random.randint(0, img1.shape[0] - self.crop_size[0])
+        x0 = np.random.randint(0, img1.shape[1] - self.crop_size[1])
+        
+        img1 = img1[y0:y0+self.crop_size[0], x0:x0+self.crop_size[1]]
+        img2 = img2[y0:y0+self.crop_size[0], x0:x0+self.crop_size[1]]
+        flow = flow[y0:y0+self.crop_size[0], x0:x0+self.crop_size[1]]
+
+        return img1, img2, flow
+
+    def __call__(self, img1, img2, flow):
+        img1, img2 = self.color_transform(img1, img2)
+        img1, img2 = self.eraser_transform(img1, img2)
+        img1, img2, flow = self.spatial_transform(img1, img2, flow)
+
+        img1 = np.ascontiguousarray(img1)
+        img2 = np.ascontiguousarray(img2)
+        flow = np.ascontiguousarray(flow)
+
+        return img1, img2, flow
+
+class SparseFlowAugmentor:
+    def __init__(self, crop_size, min_scale=-0.2, max_scale=0.5, do_flip=False):
+        # spatial augmentation params
+        self.crop_size = crop_size
+        self.min_scale = min_scale
+        self.max_scale = max_scale
+        self.spatial_aug_prob = 0.8
+        self.stretch_prob = 0.8
+        self.max_stretch = 0.2
+
+        # flip augmentation params
+        self.do_flip = do_flip
+        self.h_flip_prob = 0.5
+        self.v_flip_prob = 0.1
+
+        # photometric augmentation params
+        self.photo_aug = ColorJitter(brightness=0.3, contrast=0.3, saturation=0.3, hue=0.3/3.14)
+        self.asymmetric_color_aug_prob = 0.2
+        self.eraser_aug_prob = 0.5
+        
+    def color_transform(self, img1, img2):
+        image_stack = np.concatenate([img1, img2], axis=0)
+        image_stack = np.array(self.photo_aug(Image.fromarray(image_stack)), dtype=np.uint8)
+        img1, img2 = np.split(image_stack, 2, axis=0)
+        return img1, img2
+
+    def eraser_transform(self, img1, img2):
+        ht, wd = img1.shape[:2]
+        if np.random.rand() < self.eraser_aug_prob:
+            mean_color = np.mean(img2.reshape(-1, 3), axis=0)
+            for _ in range(np.random.randint(1, 3)):
+                x0 = np.random.randint(0, wd)
+                y0 = np.random.randint(0, ht)
+                dx = np.random.randint(50, 100)
+                dy = np.random.randint(50, 100)
+                img2[y0:y0+dy, x0:x0+dx, :] = mean_color
+
+        return img1, img2
+
+    def resize_sparse_flow_map(self, flow, valid, fx=1.0, fy=1.0):
+        ht, wd = flow.shape[:2]
+        coords = np.meshgrid(np.arange(wd), np.arange(ht))
+        coords = np.stack(coords, axis=-1)
+
+        coords = coords.reshape(-1, 2).astype(np.float32)
+        flow = flow.reshape(-1, 2).astype(np.float32)
+        valid = valid.reshape(-1).astype(np.float32)
+
+        coords0 = coords[valid>=1]
+        flow0 = flow[valid>=1]
+
+        ht1 = int(round(ht * fy))
+        wd1 = int(round(wd * fx))
+
+        coords1 = coords0 * [fx, fy]
+        flow1 = flow0 * [fx, fy]
+
+        xx = np.round(coords1[:,0]).astype(np.int32)
+        yy = np.round(coords1[:,1]).astype(np.int32)
+
+        v = (xx > 0) & (xx < wd1) & (yy > 0) & (yy < ht1)
+        xx = xx[v]
+        yy = yy[v]
+        flow1 = flow1[v]
+
+        flow_img = np.zeros([ht1, wd1, 2], dtype=np.float32)
+        valid_img = np.zeros([ht1, wd1], dtype=np.int32)
+
+        flow_img[yy, xx] = flow1
+        valid_img[yy, xx] = 1
+
+        return flow_img, valid_img
+
+    def spatial_transform(self, img1, img2, flow, valid):
+        # randomly sample scale
+
+        ht, wd = img1.shape[:2]
+        min_scale = np.maximum(
+            (self.crop_size[0] + 1) / float(ht), 
+            (self.crop_size[1] + 1) / float(wd))
+
+        scale = 2 ** np.random.uniform(self.min_scale, self.max_scale)
+        scale_x = np.clip(scale, min_scale, None)
+        scale_y = np.clip(scale, min_scale, None)
+
+        if np.random.rand() < self.spatial_aug_prob:
+            # rescale the images
+            img1 = cv2.resize(img1, None, fx=scale_x, fy=scale_y, interpolation=cv2.INTER_LINEAR)
+            img2 = cv2.resize(img2, None, fx=scale_x, fy=scale_y, interpolation=cv2.INTER_LINEAR)
+            flow, valid = self.resize_sparse_flow_map(flow, valid, fx=scale_x, fy=scale_y)
+
+        if self.do_flip:
+            if np.random.rand() < 0.5: # h-flip
+                img1 = img1[:, ::-1]
+                img2 = img2[:, ::-1]
+                flow = flow[:, ::-1] * [-1.0, 1.0]
+                valid = valid[:, ::-1]
+
+        margin_y = 20
+        margin_x = 50
+
+        y0 = np.random.randint(0, img1.shape[0] - self.crop_size[0] + margin_y)
+        x0 = np.random.randint(-margin_x, img1.shape[1] - self.crop_size[1] + margin_x)
+
+        y0 = np.clip(y0, 0, img1.shape[0] - self.crop_size[0])
+        x0 = np.clip(x0, 0, img1.shape[1] - self.crop_size[1])
+
+        img1 = img1[y0:y0+self.crop_size[0], x0:x0+self.crop_size[1]]
+        img2 = img2[y0:y0+self.crop_size[0], x0:x0+self.crop_size[1]]
+        flow = flow[y0:y0+self.crop_size[0], x0:x0+self.crop_size[1]]
+        valid = valid[y0:y0+self.crop_size[0], x0:x0+self.crop_size[1]]
+        return img1, img2, flow, valid
+
+
+    def __call__(self, img1, img2, flow, valid):
+        img1, img2 = self.color_transform(img1, img2)
+        img1, img2 = self.eraser_transform(img1, img2)
+        img1, img2, flow, valid = self.spatial_transform(img1, img2, flow, valid)
+
+        img1 = np.ascontiguousarray(img1)
+        img2 = np.ascontiguousarray(img2)
+        flow = np.ascontiguousarray(flow)
+        valid = np.ascontiguousarray(valid)
+
+        return img1, img2, flow, valid

SD-CN-Animation/RAFT/utils/flow_viz.py

@@ -0,0 +1,132 @@
+# Flow visualization code used from https://github.com/tomrunia/OpticalFlow_Visualization
+
+
+# MIT License
+#
+# Copyright (c) 2018 Tom Runia
+#
+# Permission is hereby granted, free of charge, to any person obtaining a copy
+# of this software and associated documentation files (the "Software"), to deal
+# in the Software without restriction, including without limitation the rights
+# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+# copies of the Software, and to permit persons to whom the Software is
+# furnished to do so, subject to conditions.
+#
+# Author: Tom Runia
+# Date Created: 2018-08-03
+
+import numpy as np
+
+def make_colorwheel():
+    """
+    Generates a color wheel for optical flow visualization as presented in:
+        Baker et al. "A Database and Evaluation Methodology for Optical Flow" (ICCV, 2007)
+        URL: http://vision.middlebury.edu/flow/flowEval-iccv07.pdf
+
+    Code follows the original C++ source code of Daniel Scharstein.
+    Code follows the the Matlab source code of Deqing Sun.
+
+    Returns:
+        np.ndarray: Color wheel
+    """
+
+    RY = 15
+    YG = 6
+    GC = 4
+    CB = 11
+    BM = 13
+    MR = 6
+
+    ncols = RY + YG + GC + CB + BM + MR
+    colorwheel = np.zeros((ncols, 3))
+    col = 0
+
+    # RY
+    colorwheel[0:RY, 0] = 255
+    colorwheel[0:RY, 1] = np.floor(255*np.arange(0,RY)/RY)
+    col = col+RY
+    # YG
+    colorwheel[col:col+YG, 0] = 255 - np.floor(255*np.arange(0,YG)/YG)
+    colorwheel[col:col+YG, 1] = 255
+    col = col+YG
+    # GC
+    colorwheel[col:col+GC, 1] = 255
+    colorwheel[col:col+GC, 2] = np.floor(255*np.arange(0,GC)/GC)
+    col = col+GC
+    # CB
+    colorwheel[col:col+CB, 1] = 255 - np.floor(255*np.arange(CB)/CB)
+    colorwheel[col:col+CB, 2] = 255
+    col = col+CB
+    # BM
+    colorwheel[col:col+BM, 2] = 255
+    colorwheel[col:col+BM, 0] = np.floor(255*np.arange(0,BM)/BM)
+    col = col+BM
+    # MR
+    colorwheel[col:col+MR, 2] = 255 - np.floor(255*np.arange(MR)/MR)
+    colorwheel[col:col+MR, 0] = 255
+    return colorwheel
+
+
+def flow_uv_to_colors(u, v, convert_to_bgr=False):
+    """
+    Applies the flow color wheel to (possibly clipped) flow components u and v.
+
+    According to the C++ source code of Daniel Scharstein
+    According to the Matlab source code of Deqing Sun
+
+    Args:
+        u (np.ndarray): Input horizontal flow of shape [H,W]
+        v (np.ndarray): Input vertical flow of shape [H,W]
+        convert_to_bgr (bool, optional): Convert output image to BGR. Defaults to False.
+
+    Returns:
+        np.ndarray: Flow visualization image of shape [H,W,3]
+    """
+    flow_image = np.zeros((u.shape[0], u.shape[1], 3), np.uint8)
+    colorwheel = make_colorwheel()  # shape [55x3]
+    ncols = colorwheel.shape[0]
+    rad = np.sqrt(np.square(u) + np.square(v))
+    a = np.arctan2(-v, -u)/np.pi
+    fk = (a+1) / 2*(ncols-1)
+    k0 = np.floor(fk).astype(np.int32)
+    k1 = k0 + 1
+    k1[k1 == ncols] = 0
+    f = fk - k0
+    for i in range(colorwheel.shape[1]):
+        tmp = colorwheel[:,i]
+        col0 = tmp[k0] / 255.0
+        col1 = tmp[k1] / 255.0
+        col = (1-f)*col0 + f*col1
+        idx = (rad <= 1)
+        col[idx]  = 1 - rad[idx] * (1-col[idx])
+        col[~idx] = col[~idx] * 0.75   # out of range
+        # Note the 2-i => BGR instead of RGB
+        ch_idx = 2-i if convert_to_bgr else i
+        flow_image[:,:,ch_idx] = np.floor(255 * col)
+    return flow_image
+
+
+def flow_to_image(flow_uv, clip_flow=None, convert_to_bgr=False):
+    """
+    Expects a two dimensional flow image of shape.
+
+    Args:
+        flow_uv (np.ndarray): Flow UV image of shape [H,W,2]
+        clip_flow (float, optional): Clip maximum of flow values. Defaults to None.
+        convert_to_bgr (bool, optional): Convert output image to BGR. Defaults to False.
+
+    Returns:
+        np.ndarray: Flow visualization image of shape [H,W,3]
+    """
+    assert flow_uv.ndim == 3, 'input flow must have three dimensions'
+    assert flow_uv.shape[2] == 2, 'input flow must have shape [H,W,2]'
+    if clip_flow is not None:
+        flow_uv = np.clip(flow_uv, 0, clip_flow)
+    u = flow_uv[:,:,0]
+    v = flow_uv[:,:,1]
+    rad = np.sqrt(np.square(u) + np.square(v))
+    rad_max = np.max(rad)
+    epsilon = 1e-5
+    u = u / (rad_max + epsilon)
+    v = v / (rad_max + epsilon)
+    return flow_uv_to_colors(u, v, convert_to_bgr)

SD-CN-Animation/RAFT/utils/frame_utils.py

@@ -0,0 +1,137 @@
+import numpy as np
+from PIL import Image
+from os.path import *
+import re
+
+import cv2
+cv2.setNumThreads(0)
+cv2.ocl.setUseOpenCL(False)
+
+TAG_CHAR = np.array([202021.25], np.float32)
+
+def readFlow(fn):
+    """ Read .flo file in Middlebury format"""
+    # Code adapted from:
+    # http://stackoverflow.com/questions/28013200/reading-middlebury-flow-files-with-python-bytes-array-numpy
+
+    # WARNING: this will work on little-endian architectures (eg Intel x86) only!
+    # print 'fn = %s'%(fn)
+    with open(fn, 'rb') as f:
+        magic = np.fromfile(f, np.float32, count=1)
+        if 202021.25 != magic:
+            print('Magic number incorrect. Invalid .flo file')
+            return None
+        else:
+            w = np.fromfile(f, np.int32, count=1)
+            h = np.fromfile(f, np.int32, count=1)
+            # print 'Reading %d x %d flo file\n' % (w, h)
+            data = np.fromfile(f, np.float32, count=2*int(w)*int(h))
+            # Reshape data into 3D array (columns, rows, bands)
+            # The reshape here is for visualization, the original code is (w,h,2)
+            return np.resize(data, (int(h), int(w), 2))
+
+def readPFM(file):
+    file = open(file, 'rb')
+
+    color = None
+    width = None
+    height = None
+    scale = None
+    endian = None
+
+    header = file.readline().rstrip()
+    if header == b'PF':
+        color = True
+    elif header == b'Pf':
+        color = False
+    else:
+        raise Exception('Not a PFM file.')
+
+    dim_match = re.match(rb'^(\d+)\s(\d+)\s$', file.readline())
+    if dim_match:
+        width, height = map(int, dim_match.groups())
+    else:
+        raise Exception('Malformed PFM header.')
+
+    scale = float(file.readline().rstrip())
+    if scale < 0: # little-endian
+        endian = '<'
+        scale = -scale
+    else:
+        endian = '>' # big-endian
+
+    data = np.fromfile(file, endian + 'f')
+    shape = (height, width, 3) if color else (height, width)
+
+    data = np.reshape(data, shape)
+    data = np.flipud(data)
+    return data
+
+def writeFlow(filename,uv,v=None):
+    """ Write optical flow to file.
+    
+    If v is None, uv is assumed to contain both u and v channels,
+    stacked in depth.
+    Original code by Deqing Sun, adapted from Daniel Scharstein.
+    """
+    nBands = 2
+
+    if v is None:
+        assert(uv.ndim == 3)
+        assert(uv.shape[2] == 2)
+        u = uv[:,:,0]
+        v = uv[:,:,1]
+    else:
+        u = uv
+
+    assert(u.shape == v.shape)
+    height,width = u.shape
+    f = open(filename,'wb')
+    # write the header
+    f.write(TAG_CHAR)
+    np.array(width).astype(np.int32).tofile(f)
+    np.array(height).astype(np.int32).tofile(f)
+    # arrange into matrix form
+    tmp = np.zeros((height, width*nBands))
+    tmp[:,np.arange(width)*2] = u
+    tmp[:,np.arange(width)*2 + 1] = v
+    tmp.astype(np.float32).tofile(f)
+    f.close()
+
+
+def readFlowKITTI(filename):
+    flow = cv2.imread(filename, cv2.IMREAD_ANYDEPTH|cv2.IMREAD_COLOR)
+    flow = flow[:,:,::-1].astype(np.float32)
+    flow, valid = flow[:, :, :2], flow[:, :, 2]
+    flow = (flow - 2**15) / 64.0
+    return flow, valid
+
+def readDispKITTI(filename):
+    disp = cv2.imread(filename, cv2.IMREAD_ANYDEPTH) / 256.0
+    valid = disp > 0.0
+    flow = np.stack([-disp, np.zeros_like(disp)], -1)
+    return flow, valid
+
+
+def writeFlowKITTI(filename, uv):
+    uv = 64.0 * uv + 2**15
+    valid = np.ones([uv.shape[0], uv.shape[1], 1])
+    uv = np.concatenate([uv, valid], axis=-1).astype(np.uint16)
+    cv2.imwrite(filename, uv[..., ::-1])
+    
+
+def read_gen(file_name, pil=False):
+    ext = splitext(file_name)[-1]
+    if ext == '.png' or ext == '.jpeg' or ext == '.ppm' or ext == '.jpg':
+        return Image.open(file_name)
+    elif ext == '.bin' or ext == '.raw':
+        return np.load(file_name)
+    elif ext == '.flo':
+        return readFlow(file_name).astype(np.float32)
+    elif ext == '.pfm':
+        flow = readPFM(file_name).astype(np.float32)
+        if len(flow.shape) == 2:
+            return flow
+        else:
+            return flow[:, :, :-1]
+    return []

SD-CN-Animation/RAFT/utils/utils.py

@@ -0,0 +1,82 @@
+import torch
+import torch.nn.functional as F
+import numpy as np
+from scipy import interpolate
+
+
+class InputPadder:
+    """ Pads images such that dimensions are divisible by 8 """
+    def __init__(self, dims, mode='sintel'):
+        self.ht, self.wd = dims[-2:]
+        pad_ht = (((self.ht // 8) + 1) * 8 - self.ht) % 8
+        pad_wd = (((self.wd // 8) + 1) * 8 - self.wd) % 8
+        if mode == 'sintel':
+            self._pad = [pad_wd//2, pad_wd - pad_wd//2, pad_ht//2, pad_ht - pad_ht//2]
+        else:
+            self._pad = [pad_wd//2, pad_wd - pad_wd//2, 0, pad_ht]
+
+    def pad(self, *inputs):
+        return [F.pad(x, self._pad, mode='replicate') for x in inputs]
+
+    def unpad(self,x):
+        ht, wd = x.shape[-2:]
+        c = [self._pad[2], ht-self._pad[3], self._pad[0], wd-self._pad[1]]
+        return x[..., c[0]:c[1], c[2]:c[3]]
+
+def forward_interpolate(flow):
+    flow = flow.detach().cpu().numpy()
+    dx, dy = flow[0], flow[1]
+
+    ht, wd = dx.shape
+    x0, y0 = np.meshgrid(np.arange(wd), np.arange(ht))
+
+    x1 = x0 + dx
+    y1 = y0 + dy
+    
+    x1 = x1.reshape(-1)
+    y1 = y1.reshape(-1)
+    dx = dx.reshape(-1)
+    dy = dy.reshape(-1)
+
+    valid = (x1 > 0) & (x1 < wd) & (y1 > 0) & (y1 < ht)
+    x1 = x1[valid]
+    y1 = y1[valid]
+    dx = dx[valid]
+    dy = dy[valid]
+
+    flow_x = interpolate.griddata(
+        (x1, y1), dx, (x0, y0), method='nearest', fill_value=0)
+
+    flow_y = interpolate.griddata(
+        (x1, y1), dy, (x0, y0), method='nearest', fill_value=0)
+
+    flow = np.stack([flow_x, flow_y], axis=0)
+    return torch.from_numpy(flow).float()
+
+
+def bilinear_sampler(img, coords, mode='bilinear', mask=False):
+    """ Wrapper for grid_sample, uses pixel coordinates """
+    H, W = img.shape[-2:]
+    xgrid, ygrid = coords.split([1,1], dim=-1)
+    xgrid = 2*xgrid/(W-1) - 1
+    ygrid = 2*ygrid/(H-1) - 1
+
+    grid = torch.cat([xgrid, ygrid], dim=-1)
+    img = F.grid_sample(img, grid, align_corners=True)
+
+    if mask:
+        mask = (xgrid > -1) & (ygrid > -1) & (xgrid < 1) & (ygrid < 1)
+        return img, mask.float()
+
+    return img
+
+
+def coords_grid(batch, ht, wd, device):
+    coords = torch.meshgrid(torch.arange(ht, device=device), torch.arange(wd, device=device))
+    coords = torch.stack(coords[::-1], dim=0).float()
+    return coords[None].repeat(batch, 1, 1, 1)
+
+
+def upflow8(flow, mode='bilinear'):
+    new_size = (8 * flow.shape[2], 8 * flow.shape[3])
+    return  8 * F.interpolate(flow, size=new_size, mode=mode, align_corners=True)

SD-CN-Animation/examples/flower_1.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a8db0719d9f215b775ae1b5dae912a425bc010f0586b41894c14bb8ad042711e
+size 1259280

SD-CN-Animation/examples/flower_11.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7499401998e41c65471963d6cbd70568908dd83a8c957a43940df99be7c52026
+size 1328049

SD-CN-Animation/examples/girl_to_jc.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d09ded8b44f7e30d55d5d6245d9ec7fa3b95e970a8c29d2c544b6c288341e39
+size 5274033

SD-CN-Animation/examples/girl_to_wc.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd730de667b8e7ea5af2dddcf129095694349f126e06291c9b1c2bb7d49843a8
+size 5630710