Spaces:

sichaolong
/

lama-cleaner-demo

Runtime error

App Files Files Community

sichaolong commited on Mar 29, 2023

Commit

4baf7bf

•

1 Parent(s): 4f0aaa3

Upload 97 files

Browse files

This view is limited to 50 files because it contains too many changes. See raw diff

Files changed (50) hide show

README.md +4 -6
__init__.py +11 -0
app.py +53 -0
app/build/asset-manifest.json +16 -0
app/build/index.html +18 -0
app/build/static/css/main.e24c9a9b.css +1 -0
app/build/static/js/main.a2cdd7a2.js +0 -0
app/build/static/js/main.a2cdd7a2.js.LICENSE.txt +60 -0
app/build/static/js/main.ca662570.js +0 -0
app/build/static/js/main.ca662570.js.LICENSE.txt +60 -0
app/build/static/js/main.ed69b879.js +0 -0
app/build/static/js/main.ed69b879.js.LICENSE.txt +60 -0
app/build/static/media/WorkSans-Black.67c2c5a144333953880b.ttf +0 -0
app/build/static/media/WorkSans-Bold.2bea7a7f7d052c74da25.ttf +0 -0
app/build/static/media/WorkSans-Regular.bb287b894b27372d8ea7.ttf +0 -0
app/build/static/media/WorkSans-SemiBold.1e98db4eb705b586728e.ttf +0 -0
app/build/static/media/coffee-machine-lineal.ee32631219cc3986f861.gif +0 -0
benchmark.py +108 -0
const.py +80 -0
ext/__init__.py +1 -0
ext/__pycache__/__init__.cpython-38.pyc +0 -0
ext/__pycache__/__init__.cpython-39.pyc +0 -0
ext/__pycache__/image_watermark_handler.cpython-38.pyc +0 -0
ext/__pycache__/image_watermark_handler.cpython-39.pyc +0 -0
ext/image_watermark_handler.py +93 -0
ext/request_info.txt +44 -0
ext/test.py +341 -0
file_manager/__init__.py +1 -0
file_manager/__pycache__/__init__.cpython-38.pyc +0 -0
file_manager/__pycache__/__init__.cpython-39.pyc +0 -0
file_manager/__pycache__/file_manager.cpython-38.pyc +0 -0
file_manager/__pycache__/file_manager.cpython-39.pyc +0 -0
file_manager/__pycache__/storage_backends.cpython-38.pyc +0 -0
file_manager/__pycache__/storage_backends.cpython-39.pyc +0 -0
file_manager/__pycache__/utils.cpython-38.pyc +0 -0
file_manager/__pycache__/utils.cpython-39.pyc +0 -0
file_manager/file_manager.py +264 -0
file_manager/storage_backends.py +46 -0
file_manager/utils.py +67 -0
helper.py +284 -0
interactive_seg.py +203 -0
make_gif.py +125 -0
model/__init__.py +0 -0
model/__pycache__/__init__.cpython-38.pyc +0 -0
model/__pycache__/__init__.cpython-39.pyc +0 -0
model/__pycache__/base.cpython-38.pyc +0 -0
model/__pycache__/base.cpython-39.pyc +0 -0
model/__pycache__/fcf.cpython-38.pyc +0 -0
model/__pycache__/instruct_pix2pix.cpython-38.pyc +0 -0
model/__pycache__/lama.cpython-38.pyc +0 -0

README.md CHANGED Viewed

@@ -1,11 +1,9 @@
 ---
 title: Lama Cleaner Demo
-emoji: 📈
-colorFrom: blue
-colorTo: pink
-sdk: gradio
-sdk_version: 3.23.0
-app_file: app.py
 pinned: false
 license: apache-2.0
 ---

 ---
 title: Lama Cleaner Demo
+emoji: 👀
+colorFrom: purple
+colorTo: purple
+sdk: docker
 pinned: false
 license: apache-2.0
 ---

__init__.py ADDED Viewed

	@@ -0,0 +1,11 @@

+import warnings
+warnings.simplefilter("ignore", UserWarning)
+from parse_args import parse_args
+def entry_point():
+    args = parse_args()
+    # To make os.environ["XDG_CACHE_HOME"] = args.model_cache_dir works for diffusers
+    # https://github.com/huggingface/diffusers/blob/be99201a567c1ccd841dc16fb24e88f7f239c187/src/diffusers/utils/constants.py#L18
+    from lama_cleaner.server import main
+    main(args)

app.py ADDED Viewed

	@@ -0,0 +1,53 @@

+from typing import List
+from pydantic import BaseModel
+from server import main
+"""
+LAMA Cleaner是一款用于图像去噪的工具，它使用了一种称为“局部自适应均值”的算法来去除图像中的噪声。
+在使用LAMA Cleaner时，您需要提供一个掩码图像，以指示哪些区域需要去噪，哪些区域不需要去噪。
+掩码图像可以是任何格式的图像文件，例如PNG、JPEG或BMP。
+您可以使用任何图像编辑软件（例如Photoshop或GIMP）创建掩码图像。在掩码图像中，您需要使用黑色和白色来表示需要去噪和不需要去噪的区域。黑色表示需要去噪的区域，白色表示不需要去噪的区域。
+创建掩码图像的步骤如下：
+1、打开您要去噪的图像和一个空白图像。
+2、在空白图像上使用画笔工具绘制黑色和白色的区域，以指示需要去噪和不需要去噪的区域。
+3、将掩码图像保存为PNG、JPEG或BMP格式。
+4、在使用LAMA Cleaner时，将掩码图像作为输入参数传递给它。
+请注意，掩码图像的质量对去噪效果有很大影响。因此，您需要花费一些时间来创建一个准确的掩码图像，以获得最佳的去噪效果。
+输入图像：要去噪的原始图像。
+掩码图像：指示哪些区域需要去噪，哪些区域不需要去噪的掩码图像。掩码图像可以是任何格式的图像文件，例如PNG、JPEG或BMP。在掩码图像中，您需要使用黑色和白色来表示需要去噪和不需要去噪的区域。黑色表示需要去噪的区域，白色表示不需要去噪的区域。
+块大小：用于计算局部均值的块的大小。块大小越大，去噪效果越好，但计算时间也会增加。
+块步长：用于计算局部均值的块的步长。步长越小，去噪效果越好，但计算时间也会增加。
+搜索窗口大小：用于搜索最佳匹配块的窗口大小。窗口大小越大，去噪效果越好，但计算时间也会增加。
+相似度阈值：用于确定最佳匹配块的相似度阈值。相似度阈值越小，去噪效果越好，但计算时间也会增加。
+去噪强度：控制去噪的强度。去噪强度越大，去噪效果越好，但可能会导致图像细节的丢失。
+请注意，这些参数的最佳值取决于您的图像和应用场景。您需要根据实际情况进行调整，以获得最佳的去噪效果。
+"""
+class FakeArgs(BaseModel):
+    host: str = "127.0.0.1"
+    port: int = 7860
+    model: str = 'lama' # 使用的模型
+    hf_access_token: str = ""
+    sd_disable_nsfw: bool = False # 禁用稳定扩散NSFW检查器。
+    sd_cpu_textencoder: bool = True # 始终在CPU上运行稳定扩散TextEncoder模型。
+    sd_run_local: bool = False
+    sd_enable_xformers: bool = False
+    local_files_only: bool = False
+    cpu_offload: bool = False
+    device: str = "cpu" # CUDA /中央处理器/多处理器
+    gui: bool = False
+    gui_size: List[int] = [1000, 1000]
+    input: str = ''
+    disable_model_switch: bool = False
+    debug: bool = False
+    no_half: bool = False
+    disable_nsfw: bool = False
+    enable_xformers: bool = False
+    model_dir: str = ""
+    output_dir: str = "resources" # 自己指定文件上传的位置
+if __name__ == "__main__":
+    main(FakeArgs())

app/build/asset-manifest.json ADDED Viewed

	@@ -0,0 +1,16 @@

+{
+  "files": {
+    "main.css": "/static/css/main.e24c9a9b.css",
+    "main.js": "/static/js/main.ca662570.js",
+    "static/media/coffee-machine-lineal.gif": "/static/media/coffee-machine-lineal.ee32631219cc3986f861.gif",
+    "static/media/WorkSans-SemiBold.ttf": "/static/media/WorkSans-SemiBold.1e98db4eb705b586728e.ttf",
+    "static/media/WorkSans-Bold.ttf": "/static/media/WorkSans-Bold.2bea7a7f7d052c74da25.ttf",
+    "static/media/WorkSans-Regular.ttf": "/static/media/WorkSans-Regular.bb287b894b27372d8ea7.ttf",
+    "static/media/WorkSans-Black.ttf": "/static/media/WorkSans-Black.67c2c5a144333953880b.ttf",
+    "index.html": "/index.html"
+  },
+  "entrypoints": [
+    "static/css/main.e24c9a9b.css",
+    "static/js/main.ca662570.js"
+  ]
+}

app/build/index.html ADDED Viewed

	@@ -0,0 +1,18 @@

+<!doctype html>
+<html lang="en">
+<head>
+    <meta http-equiv="Cache-Control" content="no-cache, no-store, must-revalidate"/>
+    <meta http-equiv="Pragma" content="no-cache"/>
+    <meta http-equiv="Expires" content="0"/>
+    <meta charset="utf-8"/>
+    <meta name="viewport" content="width=device-width,initial-scale=1,maximum-scale=1,user-scalable=0"/>
+    <meta name="theme-color" content="#ffffff"/>
+    <title>lama-cleaner - Image inpainting powered by SOTA AI model</title>
+    <script defer="defer" src="/static/js/main.ca662570.js"></script>
+    <link href="/static/css/main.e24c9a9b.css" rel="stylesheet">
+</head>
+<body>
+<noscript>You need to enable JavaScript to run this app.</noscript>
+<div id="root"></div>
+</body>
+</html>

app/build/static/css/main.e24c9a9b.css ADDED Viewed

	@@ -0,0 +1 @@

+ :root{--blackA1:rgba(0,0,0,.012);--blackA2:rgba(0,0,0,.027);--blackA3:rgba(0,0,0,.047);--blackA4:rgba(0,0,0,.071);--blackA5:rgba(0,0,0,.09);--blackA6:rgba(0,0,0,.114);--blackA7:rgba(0,0,0,.141);--blackA8:rgba(0,0,0,.22);--blackA9:rgba(0,0,0,.439);--blackA10:rgba(0,0,0,.478);--blackA11:rgba(0,0,0,.565);--blackA12:rgba(0,0,0,.91);--mauve1:#fdfcfd;--mauve2:#f9f8f9;--mauve3:#f4f2f4;--mauve4:#eeedef;--mauve5:#e9e8ea;--mauve6:#e4e2e4;--mauve7:#dcdbdd;--mauve8:#c8c7cb;--mauve9:#908e96;--mauve10:#86848d;--mauve11:#6f6e77;--mauve12:#1a1523;--violet1:#fdfcfe;--violet2:#fbfaff;--violet3:#f5f2ff;--violet4:#ede9fe;--violet5:#e4defc;--violet6:#d7cff9;--violet7:#c4b8f3;--violet8:#aa99ec;--violet9:#6e56cf;--violet10:#644fc1;--violet11:#5746af;--violet12:#20134b;--page-bg:#fff;--page-bg-light:hsla(0,0%,100%,.5);--page-text-color:#040404;--yellow-accent:#fc0;--yellow-accent-light:#ffcc0055;--link-color:#000;--border-color:#eff1f4;--border-color-light:hsla(240,9%,43%,.5);--tooltip-bg:#e6e6ea;--tooltip-text-color:#000;--error-color:#ef4444;--success-color:#10b981;--editor-toolkit-bg:hsla(0,0%,100%,.5);--editor-options-bg:#e6e6ea;--options-text-color:var(--page-text-color);--editor-size-border-color:var(--border-color);--editor-toolkit-panel-border:0;--modal-bg:var(--page-bg);--modal-text-color:#000;--modal-hotkey-border-color:#000;--model-mask-bg:rgba(209,213,219,.4);--text-color:#040404;--text-color-gray:#6b6f76;--btn-text-color:var(--text-color);--btn-text-hover-color:#040404;--btn-border-color:#646478;--btn-primary-hover-bg:var(--yellow-accent);--animation-pulsing-bg:hsla(0,0%,100%,.5);--switch-root-background-color:#dfe1e4;--switch-thumb-color:var(--page-bg);--switch-thumb-checked-color:var(--page-bg);--slider-background-color:var(--switch-root-background-color);--tooltip-bg:var(--page-bg);--badge-background-color:#f1f3f5;--badge-color:#687076;--box-shadow:inset 0 0.5px hsla(0,0%,100%,.1),inset 0 1px 5px #f8f9fa,0px 0px 0px 0.5px #c1c8cd,0px 2px 1px -1px #c1c8cd,0 1px #c1c8cd;--croper-bg:rgba(0,0,0,.5);--tabs-active-color:#f0f3f9}@font-face{font-family:WorkSans;src:url(/static/media/WorkSans-Regular.bb287b894b27372d8ea7.ttf)}@font-face{font-family:WorkSans-Semibold;src:url(/static/media/WorkSans-SemiBold.1e98db4eb705b586728e.ttf)}@font-face{font-family:WorkSans-Bold;src:url(/static/media/WorkSans-Bold.2bea7a7f7d052c74da25.ttf)}@font-face{font-family:WorkSans-Black;src:url(/static/media/WorkSans-Black.67c2c5a144333953880b.ttf)}[data-theme=dark]{--page-bg:#040404;--page-bg-light:#04040488;--page-text-color:#f9f9f9;--yellow-accent:#fc0;--yellow-accent-light:#ffcc0055;--link-color:var(--yellow-accent);--border-color:#1e1e1e;--border-color-light:#666;--tooltip-bg:#212121;--tooltip-text-color:#d2d2d2;--editor-toolkit-bg:rgba(0,0,0,.5);--editor-options-bg:#212121;--options-text-color:var(--page-text-color);--editor-size-border-color:var(--yellow-accent);--editor-toolkit-panel-border:1px solid hsla(240,9%,43%,.4);--modal-bg:var(--page-bg);--modal-text-color:var(--page-text-color);--modal-hotkey-border-color:var(--page-text-color);--model-mask-bg:rgba(76,76,87,.4);--text-color:#fff;--text-color-gray:#c3c4c6;--btn-text-color:var(--text-color);--btn-text-hover-color:var(--page-bg);--btn-border-color:var(--yellow-accent);--btn-primary-hover-bg:var(--yellow-accent);--animation-pulsing-bg:#f0f0ff;--switch-root-background-color:#3c3f44;--switch-thumb-color:#1f2023;--switch-thumb-checked-color:#fff;--slider-background-color:var(--switch-root-background-color);--badge-background-color:#202425;--badge-color:#9ba1a6;--box-shadow:inset 0 0.5px hsla(0,0%,100%,.1),inset 0 1px 5px #1a1d1e,0px 0px 0px 0.5px #4c5155,0px 2px 1px -1px #4c5155,0 1px #4c5155;--croper-bg:rgba(0,0,0,.5);--tabs-active-color:#272831}@supports (color:hsl(0 0% 0%/0)){[data-theme=dark]{--tooltip-bg:#202425}}@-webkit-keyframes pulsing{0%{opacity:1}50%{background-color:hsla(0,0%,100%,.5);background-color:var(--animation-pulsing-bg);opacity:.75}to{opacity:1}}@keyframes pulsing{0%{opacity:1}50%{background-color:hsla(0,0%,100%,.5);background-color:var(--animation-pulsing-bg);opacity:.75}to{opacity:1}}@-webkit-keyframes opacityReveal{0%{opacity:0}to{opacity:1}}@keyframes opacityReveal{0%{opacity:0}to{opacity:1}}@-webkit-keyframes slideDown{0%{-webkit-transform:translateY(-100%);transform:translateY(-100%)}to{-webkit-transform:translateY(0);transform:translateY(0)}}@keyframes slideDown{0%{-webkit-transform:translateY(-100%);transform:translateY(-100%)}to{-webkit-transform:translateY(0);transform:translateY(0)}}@-webkit-keyframes slideUp{0%{-webkit-transform:translateY(100%);transform:translateY(100%)}to{-webkit-transform:translateY(0);transform:translateY(0)}}@keyframes slideUp{0%{-webkit-transform:translateY(100%);transform:translateY(100%)}to{-webkit-transform:translateY(0);transform:translateY(0)}}@-webkit-keyframes slideIn{0%{-webkit-transform:translateX(calc(100% + 25px));transform:translateX(calc(100% + 25px))}to{-webkit-transform:translateX(0);transform:translateX(0)}}@keyframes slideIn{0%{-webkit-transform:translateX(calc(100% + 25px));transform:translateX(calc(100% + 25px))}to{-webkit-transform:translateX(0);transform:translateX(0)}}@-webkit-keyframes spin{0%{-webkit-transform:rotate(0deg);transform:rotate(0deg)}to{-webkit-transform:rotate(1turn);transform:rotate(1turn)}}@keyframes spin{0%{-webkit-transform:rotate(0deg);transform:rotate(0deg)}to{-webkit-transform:rotate(1turn);transform:rotate(1turn)}}@-webkit-keyframes slideUpAndFade{0%{opacity:0;-webkit-transform:translateY(2px);transform:translateY(2px)}to{opacity:1;-webkit-transform:translateY(0);transform:translateY(0)}}@keyframes slideUpAndFade{0%{opacity:0;-webkit-transform:translateY(2px);transform:translateY(2px)}to{opacity:1;-webkit-transform:translateY(0);transform:translateY(0)}}@-webkit-keyframes slideDownAndFade{0%{opacity:0;-webkit-transform:translateY(-2px);transform:translateY(-2px)}to{opacity:1;-webkit-transform:translateY(0);transform:translateY(0)}}@keyframes slideDownAndFade{0%{opacity:0;-webkit-transform:translateY(-2px);transform:translateY(-2px)}to{opacity:1;-webkit-transform:translateY(0);transform:translateY(0)}}.lama-cleaner{background-color:#fff;background-color:var(--page-bg);color:#040404;color:var(--page-text-color);display:grid;grid-template-areas:"main-content";height:100vh;transition-duration:.2s;transition-property:background-color,color;transition-timing-function:repeat(2,ease-out);width:100vw}a{color:inherit;text-decoration:inherit}input:disabled{color:#6b6f76;color:var(--text-color-gray)}.editor-container{align-items:center;display:flex;height:100vh;justify-content:center;width:100vw}.react-transform-wrapper{display:grid!important;height:100%!important;width:100%!important}.editor-canvas-container{grid-row-gap:1rem;display:grid;grid-template-areas:"editor-content";row-gap:1rem}.editor-canvas{grid-area:editor-content;z-index:2}.original-image-container{display:grid;grid-area:editor-content;grid-template-areas:"original-image-content";pointer-events:none}.original-image-container img{grid-area:original-image-content}.original-image-container .editor-slider{background-color:#fc0;background-color:var(--yellow-accent);grid-area:original-image-content;height:100%;justify-self:end;transition:all .3s cubic-bezier(.4,0,.2,1);width:6px;z-index:2}.editor-canvas-loading{-webkit-animation:pulsing .75s infinite;animation:pulsing .75s infinite;pointer-events:none}.editor-toolkit-panel{align-items:center;-webkit-animation:slideUp .2s ease-out;animation:slideUp .2s ease-out;-webkit-backdrop-filter:blur(12px);backdrop-filter:blur(12px);background-color:hsla(0,0%,100%,.5);background-color:var(--page-bg-light);border:0;border:var(--editor-toolkit-panel-border);border-radius:3rem;bottom:.5rem;box-shadow:0 0 0 1px rgba(0,0,0,.102),0 3px 16px rgba(0,0,0,.078),0 2px 6px 1px rgba(0,0,0,.09);display:flex;gap:16px;justify-content:center;padding:.6rem 32px;position:fixed}@media screen and (max-width:767px){.editor-toolkit-panel{grid-template-areas:"toolkit-size-selector toolkit-size-selector" "toolkit-brush-slider toolkit-brush-slider" "toolkit-btns toolkit-btns";justify-items:center;padding:1rem 2rem;row-gap:2rem}}.editor-toolkit-panel .eyeicon-active{background-color:#fc0;background-color:var(--yellow-accent);color:#040404;color:var(--btn-text-hover-color)}.editor-brush-slider{grid-column-gap:1rem;align-items:center;-webkit-column-gap:1rem;column-gap:1rem;display:grid;grid-area:toolkit-brush-slider;grid-template-columns:repeat(2,-webkit-max-content);grid-template-columns:repeat(2,max-content);height:-webkit-max-content;height:max-content;-webkit-user-select:none;user-select:none}.editor-brush-slider input[type=range]{-webkit-appearance:none;appearance:none;background:transparent;border-color:transparent;color:transparent;cursor:pointer;width:100%}.editor-brush-slider input[type=range]:focus{outline:none}.editor-brush-slider input[type=range]::-webkit-slider-thumb{-webkit-appearance:none;background:#fc0;background:var(--yellow-accent);border:1px solid #000;border-radius:50%;height:1.2rem;margin-top:-.5rem;width:1.2rem;z-index:2}.editor-brush-slider input[type=range]::-webkit-slider-runnable-track{background:#dfe1e4;background:var(--slider-background-color);border-radius:2rem;height:.2rem}.editor-brush-slider input[type=range]::-moz-range-track{background:#dfe1e4;background:var(--slider-background-color);border-radius:2rem}.editor-brush-slider input[type=range]::-moz-range-progress{background:#fc0;background:var(--yellow-accent)}.editor-toolkit-btns{display:flex;gap:12px}.brush-shape{background-color:rgba(255,204,0,.733);border:1px solid #fc0;border:1px solid var(--yellow-accent);border-radius:50%;pointer-events:none;position:absolute}.file-manager-modal{color:#040404;color:var(--text-color);height:90%;width:80%}.react-photo-album.react-photo-album--columns{height:80vh}.react-photo-album--photo{border:1px solid transparent;border-radius:8px;transition:visibility .25s ease-in,-webkit-transform .25s;transition:transform .25s,visibility .25s ease-in;transition:transform .25s,visibility .25s ease-in,-webkit-transform .25s;-webkit-user-select:none;user-select:none}.react-photo-album--photo:hover{border:1px solid #eff1f4;border:1px solid var(--border-color);-webkit-transform:scale(1.03);transform:scale(1.03)}.ScrollAreaRoot{--scrollbar-size:10px;border-radius:4px;overflow:hidden}.ScrollAreaViewport{border-radius:inherit;height:100%;width:100%}.ScrollAreaScrollbar{display:flex;padding:2px;touch-action:none;transition:background .16s ease-out;-webkit-user-select:none;user-select:none}.ScrollAreaScrollbar:hover{background:var(--blackA8)}.ScrollAreaScrollbar[data-orientation=vertical]{width:var(--scrollbar-size)}.ScrollAreaScrollbar[data-orientation=horizontal]{flex-direction:column;height:var(--scrollbar-size)}.ScrollAreaThumb{background:var(--mauve10);border-radius:var(--scrollbar-size);flex:1 1;position:relative}.ScrollAreaThumb:before{content:"";height:100%;left:50%;min-height:44px;min-width:44px;position:absolute;top:50%;-webkit-transform:translate(-50%,-50%);transform:translate(-50%,-50%);width:100%}.ScrollAreaCorner{background:var(--blackA8)}.file-search-input{border:1px solid #eff1f4;border:1px solid var(--border-color);border-radius:12px;height:32px;padding-left:30px;width:250px}.sort-btn-inactive svg{opacity:.5}button,fieldset,input{all:unset}.TabsRoot{align-self:flex-start;background-color:#fff;background-color:var(--page-bg);display:flex;flex-direction:column;gap:8px}.TabsList{border:1px solid #eff1f4;border:1px solid var(--border-color);border-radius:12px;flex-direction:row;gap:6px;padding:4px}.TabsList,.TabsTrigger{background-color:#fff;background-color:var(--page-bg);display:flex;justify-content:flex-start}.TabsTrigger{align-items:center;border-radius:8px;color:#000;color:var(--modal-text-color);font-family:inherit;font-size:15px;line-height:1;padding:8px;-webkit-user-select:none;user-select:none}.TabsTrigger:hover,.TabsTrigger[data-state=active]{background-color:#f0f3f9;background-color:var(--tabs-active-color)}.TabsTrigger:focus{position:relative}.TabsContent{background-color:#fff;background-color:var(--page-bg);outline:none;width:100%}.TabsContent[data-state=active]{display:flex;flex-direction:column;gap:14px}.landing-page{grid-row-gap:2rem;display:grid;grid-auto-rows:-webkit-max-content;grid-auto-rows:max-content;justify-items:center;place-self:center;row-gap:2rem}@media screen and (max-width:767px){.landing-page{padding:1rem}}.landing-page h1{font-size:1.4rem;text-align:center}@media screen and (max-width:767px){.landing-page h1{font-size:1.2rem}}.landing-page a{color:#000;color:var(--link-color)}.landing-file-selector{display:grid}header{align-items:center;-webkit-backdrop-filter:blur(12px);backdrop-filter:blur(12px);background-color:hsla(0,0%,100%,.5);background-color:var(--page-bg-light);border-bottom:1px solid hsla(240,9%,43%,.2);display:flex;height:60px;justify-content:space-between;padding:1rem 1.5rem;position:absolute;top:0;width:100%;z-index:20}.shortcuts{z-index:1}.header-icons-wrapper{gap:12px}.header-icons,.header-icons-wrapper{align-items:center;display:flex;justify-content:center;justify-self:end}.header-icons{gap:6px}.mask-preview{border:1px solid #eff1f4;border:1px solid var(--border-color);border-radius:8px;margin-left:20px;margin-top:30px;max-height:400px;max-width:400px}.prompt-wrapper{display:flex;gap:12px}.prompt-wrapper input{all:unset;border-radius:.5rem;border-width:0;min-width:600px;outline:1px solid #eff1f4;outline:1px solid var(--border-color);padding:0 .8rem}.prompt-wrapper input:focus-visible{border-width:0;outline:1px solid #fc0;outline:1px solid var(--yellow-accent)}.theme-toggle-ui{transition:all .2s ease-in;-webkit-user-select:none;user-select:none;z-index:10}.theme-toggle-ui .theme-btn{align-items:center;cursor:pointer;display:flex;justify-content:center;outline:none}.theme-toggle-ui .theme-btn svg{height:22px;width:22px}.modal-shortcuts{background-color:#fff;background-color:var(--modal-bg);box-shadow:0 0 20px rgba(0,0,40,.2);color:#000;color:var(--modal-text-color);grid-area:main-content}@media screen and (max-width:767px){.modal-shortcuts{-webkit-animation:slideDown .2s ease-out;animation:slideDown .2s ease-out;display:grid;height:auto;width:100%}}.shortcut-options{display:flex;flex-direction:row;gap:48px}.shortcut-options .shortcut-option{grid-column-gap:2rem;align-items:center;-webkit-column-gap:2rem;column-gap:2rem;display:grid;grid-template-columns:repeat(2,auto)}@media screen and (max-width:767px){.shortcut-options .shortcut-option{-webkit-column-gap:0;column-gap:0;row-gap:.6rem}}.shortcut-options .shortcut-key{background-color:#fff;background-color:var(--page-bg);border-radius:6px;box-shadow:inset 0 .5px hsla(0,0%,100%,.1),inset 0 1px 5px #f8f9fa,0 0 0 .5px #c1c8cd,0 2px 1px -1px #c1c8cd,0 1px #c1c8cd;box-shadow:var(--box-shadow);box-sizing:border-box;color:#000;color:var(--modal-text-color);font-family:inherit;font-weight:400;justify-self:end;line-height:1.5;padding-left:.5rem;padding-right:.5rem;text-shadow:0 0 1px hsla(0,0%,100%,.5);-webkit-user-select:none;user-select:none;white-space:nowrap;width:-webkit-max-content;width:max-content}@media screen and (max-width:767px){.shortcut-options .shortcut-key{padding:.2rem .4rem}}.shortcut-options .shortcut-description{font-size:.95rem;justify-self:start;text-align:left}@media screen and (max-width:767px){.shortcut-options .shortcut-description{justify-self:start;text-align:left;width:auto}}.shortcut-options-column{gap:12px;width:320px}.setting-block,.setting-block .option-desc,.shortcut-options-column{display:flex;flex-direction:column}.setting-block .option-desc{border:1px solid #eff1f4;border:1px solid var(--border-color);border-radius:.3rem;color:#6b6f76;color:var(--text-color-gray);gap:8px;margin-top:12px;padding:1rem}.setting-block .option-desc .sub-setting-block{color:#040404;color:var(--text-color)}.setting-block .option-desc svg{color:#6b6f76;color:var(--text-color-gray)}.setting-block-content{align-items:center;display:flex;gap:12rem;justify-content:space-between}.setting-block-content-v{align-items:flex-start;display:flex;flex-direction:column;gap:1rem;justify-content:flex-start}.setting-block-content-title{align-items:center;display:flex;flex-direction:row;gap:8px;justify-content:center}.setting-block-desc{color:#6b6f76;color:var(--text-color-gray);font-size:1rem;margin-top:8px}.hd-setting-block .inline-tip{color:#040404;color:var(--text-color);cursor:pointer;display:inline}.model-desc-link{border-radius:999px;color:#687076;color:var(--badge-color);display:flex;justify-items:center;padding-left:5px;padding-right:5px;text-decoration:none}.modal-setting{background-color:#fff;background-color:var(--modal-bg);box-shadow:0 0 20px rgba(0,0,40,.2);color:#000;color:var(--modal-text-color);width:680px}@media screen and (max-width:767px){.modal-setting{-webkit-animation:slideDown .2s ease-out;animation:slideDown .2s ease-out;display:grid;height:auto;margin-top:-11rem;width:100%}}.folder-path-block{display:flex;flex-direction:column;gap:12px}.folder-path{border-radius:6px;border-width:0;outline:1px solid #eff1f4;outline:1px solid var(--border-color);padding:.3rem .5rem;width:95%}.folder-path:focus-visible{border-width:0;outline:1px solid #fc0;outline:1px solid var(--yellow-accent)}.side-panel{border-color:#eff1f4;border-color:var(--border-color);border-radius:.8rem;border-style:solid;border-width:1px;padding:.3rem;position:absolute;right:1.5rem;top:68px;z-index:4}.side-panel-trigger{border:0;font-family:WorkSans,sans-serif;font-size:16px}.side-panel-content{background-color:#fff;background-color:var(--page-bg);border-color:#eff1f4;border-color:var(--border-color);border-radius:.8rem;border-style:solid;border-width:1px;color:#040404;color:var(--text-color);display:flex;flex-direction:column;font-family:WorkSans,sans-serif;font-size:14px;gap:12px;padding:1rem;position:relative;right:1.5rem;top:1rem;z-index:9}.side-panel-content .setting-block-content{gap:1rem}.negative-prompt{all:unset;border-radius:.5rem;border-width:0;max-width:200px;min-height:150px;outline:1px solid #eff1f4;outline:1px solid var(--border-color);padding:12px .8rem;width:100%}.negative-prompt:focus-visible{border-width:0;outline:1px solid #fc0;outline:1px solid var(--yellow-accent)}.negative-prompt:-webkit-input-placeholder{padding-top:10px}.negative-prompt:-moz-input-placeholder{padding-top:10px}.negative-prompt:-ms-input-placeholder{padding-top:10px}.resize-title-tile{color:#6b6f76;color:var(--text-color-gray);font-size:.5rem;width:86px}.crop-border{outline-color:#fc0;outline-color:var(--yellow-accent);outline-style:dashed}.info-bar{align-items:center;background-color:#fff;background-color:var(--page-bg);border:0;border:var(--editor-toolkit-panel-border);border-radius:9999px;box-shadow:0 0 0 1px rgba(0,0,0,.102),0 3px 16px rgba(0,0,0,.078),0 2px 6px 1px rgba(0,0,0,.09);color:#040404;color:var(--text-color);display:flex;font-size:1rem;gap:12px;justify-content:center;padding:.2rem .8rem;pointer-events:auto;position:absolute}.info-bar:hover{cursor:move}.croper-wrapper{height:100%;overflow:hidden;position:absolute;width:100%}.croper,.croper-wrapper{pointer-events:none;z-index:2}.croper{bottom:0;box-shadow:0 0 0 9999px rgba(0,0,0,.5);left:0;position:relative;right:0;top:0}.drag-bar{pointer-events:auto;position:absolute}.drag-bar.ord-top{cursor:ns-resize;height:12px;left:0;margin-top:-6px;top:0;width:100%}.drag-bar.ord-right{cursor:ew-resize;height:100%;margin-right:-6px;right:0;top:0;width:12px}.drag-bar.ord-bottom{bottom:0;cursor:ns-resize;height:12px;left:0;margin-bottom:-6px;width:100%}.drag-bar.ord-left{cursor:ew-resize;height:100%;left:0;margin-left:-6px;top:0;width:12px}.drag-handle{background-color:#ffcc0055;background-color:var(--yellow-accent-light);border:2px solid #fc0;border:2px solid var(--yellow-accent);content:"";display:block;height:12px;pointer-events:auto;position:absolute;width:12px;z-index:4}.drag-handle:hover{background-color:#fc0;background-color:var(--yellow-accent)}.drag-handle.ord-topleft{cursor:nw-resize;left:-7px;top:-7px}.drag-handle.ord-topright{cursor:ne-resize;right:-7px;top:-7px}.drag-handle.ord-bottomright{bottom:-7px;cursor:se-resize;right:-7px}.drag-handle.ord-bottomleft{bottom:-7px;cursor:sw-resize;left:-7px}.drag-handle.ord-bottom,.drag-handle.ord-top{cursor:ns-resize;left:calc(50% - 6px)}.drag-handle.ord-top{top:-7px}.drag-handle.ord-bottom{bottom:-7px}.drag-handle.ord-left,.drag-handle.ord-right{cursor:ew-resize;top:calc(50% - 6px)}.drag-handle.ord-left{left:-7px}.drag-handle.ord-right{right:-7px}.interactive-seg-wrapper{height:100%;overflow:hidden;pointer-events:none;position:absolute;width:100%;z-index:2}.interactive-seg-wrapper .click-item{border-radius:50%;height:8px;position:absolute;width:8px}.interactive-seg-wrapper .click-item-positive{background-color:rgba(21,215,121,.936);outline:6px solid rgba(98,255,179,.31)}.interactive-seg-wrapper .click-item-negative{background-color:rgba(237,49,55,.942);outline:6px solid rgba(255,89,95,.31)}.interactive-seg-confirm-actions{background-color:#fff;background-color:var(--page-bg);border-color:#eff1f4;border-color:var(--border-color);border-radius:16px;border-style:solid;border-width:1px;padding:8px;position:absolute;top:68px;z-index:5}.interactive-seg-confirm-actions .action-buttons{align-items:center;display:flex;gap:8px;justify-content:center}@-webkit-keyframes pulse{to{box-shadow:0 0 0 14px rgba(21,215,121,0)}}@keyframes pulse{to{box-shadow:0 0 0 14px rgba(21,215,121,0)}}.interactive-seg-cursor{-webkit-animation:pulse 1.5s cubic-bezier(.66,0,0,1) infinite;animation:pulse 1.5s cubic-bezier(.66,0,0,1) infinite;background-color:rgba(21,215,121,.936);border-radius:50%;box-shadow:0 0 0 0 rgba(21,215,121,.936);color:rgba(234,255,240,.98);height:20px;pointer-events:none;position:absolute;width:20px}.file-select-label{border:2px dashed #eff1f4;border:2px dashed var(--border-color);border-radius:.5rem;cursor:pointer;display:grid;min-width:600px}@media screen and (max-width:767px){.file-select-label{min-width:300px}}.file-select-label .file-select-label-hover,.file-select-label:hover{background-color:#fc0;background-color:var(--yellow-accent);color:#000}.file-select-container{display:grid;height:100%;padding:4rem;width:100%}.file-select-container input{display:none}.file-select-message{font-family:WorkSans;text-align:center}.btn-primary{grid-column-gap:1rem;background-color:#fff;background-color:var(--page-bg);border-radius:.5rem;color:#040404;color:var(--btn-text-color);-webkit-column-gap:1rem;column-gap:1rem;cursor:pointer;display:grid;font-family:WorkSans,sans-serif;grid-auto-flow:column;padding:.5rem;place-items:center;width:-webkit-max-content;width:max-content;z-index:1}.btn-primary:hover{background-color:#fc0;background-color:var(--btn-primary-hover-bg);color:#040404;color:var(--btn-text-hover-color)}.btn-primary svg{height:auto;width:20px}.btn-primary-disabled{background-color:#fff;background-color:var(--page-bg);opacity:.5;pointer-events:none}.btn-border{border-color:#646478;border-color:var(--btn-border-color);border-style:solid;border-width:1px}.modal-mask{-webkit-backdrop-filter:blur(12px);backdrop-filter:blur(12px);background-color:rgba(209,213,219,.4);background-color:var(--model-mask-bg);inset:0;position:fixed;z-index:9998}@media(prefers-reduced-motion:no-preference){.modal-mask{-webkit-animation:opacityReveal .15s cubic-bezier(.16,1,.3,1) forwards;animation:opacityReveal .15s cubic-bezier(.16,1,.3,1) forwards}}@-webkit-keyframes contentShow{0%{opacity:0;-webkit-transform:translate(-50%,-48%) scale(.96);transform:translate(-50%,-48%) scale(.96)}to{opacity:1;-webkit-transform:translate(-50%,-50%) scale(1);transform:translate(-50%,-50%) scale(1)}}@keyframes contentShow{0%{opacity:0;-webkit-transform:translate(-50%,-48%) scale(.96);transform:translate(-50%,-48%) scale(.96)}to{opacity:1;-webkit-transform:translate(-50%,-50%) scale(1);transform:translate(-50%,-50%) scale(1)}}.modal{background-color:#fff;background-color:var(--page-bg);border-radius:.95rem;display:flex;flex-direction:column;gap:16px;left:50%;padding:25px;place-self:center;position:fixed;top:50%;-webkit-transform:translate(-50%,-50%);transform:translate(-50%,-50%);z-index:9999}.modal:focus{outline:none}.modal .modal-header{align-items:center;display:grid;grid-template-columns:repeat(2,auto)}.modal .modal-header .btn-primary{justify-self:end}@media(prefers-reduced-motion:no-preference){.modal{-webkit-animation:contentShow .15s cubic-bezier(.16,1,.3,1) forwards;animation:contentShow .15s cubic-bezier(.16,1,.3,1) forwards}}.select-trigger{all:unset;align-items:center;background-color:#fff;background-color:var(--page-bg);border:1px solid #eff1f4;border:1px solid var(--border-color);border-radius:.5rem;color:#040404;color:var(--options-text-color);display:inline-flex;gap:8px;height:32px;justify-content:space-between;padding:0 .8rem}.select-trigger svg{height:1rem;margin-top:.25rem;width:1rem}.select-trigger:hover{border-color:#fc0;border-color:var(--yellow-accent)}.select-trigger:disabled{border-color:#eff1f4;border-color:var(--border-color);color:#eff1f4;color:var(--border-color)}.select-content{background-color:#fff;background-color:var(--page-bg);border-radius:.5rem;overflow:hidden}.select-viewport{border:1px solid #eff1f4;border:1px solid var(--border-color);border-radius:.5rem;padding:5px}.select-item{all:unset;align-items:center;background-color:#fff;background-color:var(--page-bg);border-radius:.5rem;color:#040404;color:var(--options-text-color);display:flex;padding:6px 6px 6px 25px;position:relative;-webkit-user-select:none;user-select:none}.select-item:focus{background-color:#fc0;background-color:var(--yellow-accent);color:#040404;color:var(--btn-text-hover-color)}.select-item-indicator{align-items:center;display:inline-flex;justify-content:center;left:0;padding-right:4px;position:absolute;width:25px}.switch-root{-webkit-tap-highlight-color:rgba(0,0,0,0);all:"unset";background-color:#dfe1e4;background-color:var(--switch-root-background-color);border:none;border-radius:9999px;height:25px;position:relative;transition:background-color .1s;width:42px}.switch-root:focus-visible{outline:none}.switch-root[data-state=checked]{background-color:#fc0;background-color:var(--yellow-accent)}.switch-thumb{background-color:#fff;background-color:var(--switch-thumb-color);border-radius:9999px;display:block;height:17px;-webkit-transform:translateX(4px);transform:translateX(4px);transition:-webkit-transform .1s;transition:transform .1s;transition:transform .1s,-webkit-transform .1s;width:17px;will-change:transform}.switch-thumb[data-state=checked]{background-color:#fff;background-color:var(--switch-thumb-checked-color);outline:1px solid hsla(240,9%,43%,.5);-webkit-transform:translateX(21px);transform:translateX(21px)}.number-input{all:unset;border-radius:.5rem;flex:1 0 auto;height:32px;outline:1px solid #eff1f4;outline:1px solid var(--border-color);padding:0 .8rem;text-align:right}.number-input:focus-visible{outline:1px solid #fc0;outline:1px solid var(--yellow-accent)}.number-input:disabled{color:#eff1f4;color:var(--border-color)}.toast-viewpoint{bottom:48px;display:flex;flex-direction:row;gap:10px;margin:0;max-width:100vw;padding:25px;position:fixed;right:1.5rem;z-index:999999}.toast-viewpoint:focus-visible{outline:none}.toast-root{align-items:center;background-color:#fff;background-color:var(--page-bg);border:1px solid hsla(240,9%,43%,.5);border:1px solid var(--border-color-light);border-radius:.6rem;display:flex;gap:12px;padding:15px}.toast-root[data-state=open]{-webkit-animation:slideIn .15s cubic-bezier(.16,1,.3,1);animation:slideIn .15s cubic-bezier(.16,1,.3,1)}.toast-root[data-state=close]{-webkit-animation:opacityReveal .1s ease-in forwards;animation:opacityReveal .1s ease-in forwards}.toast-root[data-state=cancel]{-webkit-animation:transform .1s ease-out;animation:transform .1s ease-out;-webkit-transform:translateX(0);transform:translateX(0)}.toast-root.error{border:1px solid #ef4444;border:1px solid var(--error-color)}.toast-root.success{border:1px solid #10b981;border:1px solid var(--success-color)}.error-icon{color:#ef4444;color:var(--error-color);height:24px;width:24px}.success-icon{color:#10b981;color:var(--success-color);height:24px;width:24px}.loading-icon{-webkit-animation-duration:1.5s;animation-duration:1.5s;-webkit-animation-iteration-count:infinite;animation-iteration-count:infinite;-webkit-animation-name:spin;animation-name:spin;-webkit-animation-timing-function:linear;animation-timing-function:linear;-webkit-transform-origin:center center;transform-origin:center center}.loading-icon,.toast-desc,.toast-icon{align-items:center;display:flex}.toast-desc{color:#040404;color:var(--text-color);margin:0;min-width:240px}.tooltip-trigger{align-items:center;display:flex;justify-content:center}.tooltip-content{background-color:#fff;background-color:var(--tooltip-bg);border-radius:4px;box-shadow:0 10px 38px -10px rgba(14,18,22,.35),0 10px 20px -15px rgba(14,18,22,.2);color:#000;color:var(--tooltip-text-color);padding:10px 15px}@media(prefers-reduced-motion:no-preference){.tooltip-content{-webkit-animation-duration:.4s;animation-duration:.4s;-webkit-animation-fill-mode:forwards;animation-fill-mode:forwards;-webkit-animation-timing-function:cubic-bezier(.16,1,.3,1);animation-timing-function:cubic-bezier(.16,1,.3,1);will-change:transform,opacity}.tooltip-content[data-state=delayed-open][data-side=top]{-webkit-animation-name:slideDownAndFade;animation-name:slideDownAndFade}.tooltip-content[data-state=delayed-open][data-side=bottom]{-webkit-animation-name:slideUpAndFade;animation-name:slideUpAndFade}}.tooltip-arrow{fill:#fff;fill:var(--tooltip-bg)}*,:after,:before{box-sizing:border-box;margin:0;padding:0}body,html{font-family:WorkSans,sans-serif}

app/build/static/js/main.a2cdd7a2.js ADDED Viewed

The diff for this file is too large to render. See raw diff

app/build/static/js/main.a2cdd7a2.js.LICENSE.txt ADDED Viewed

	@@ -0,0 +1,60 @@

+/*
+object-assign
+(c) Sindre Sorhus
+@license MIT
+*/
+/*! regenerator-runtime -- Copyright (c) 2014-present, Facebook, Inc. -- license (MIT): https://github.com/facebook/regenerator/blob/main/LICENSE */
+/**
+ * @license
+ * Lodash <https://lodash.com/>
+ * Copyright OpenJS Foundation and other contributors <https://openjsf.org/>
+ * Released under MIT license <https://lodash.com/license>
+ * Based on Underscore.js 1.8.3 <http://underscorejs.org/LICENSE>
+ * Copyright Jeremy Ashkenas, DocumentCloud and Investigative Reporters & Editors
+ */
+/** @license React v0.20.2
+ * scheduler.production.min.js
+ *
+ * Copyright (c) Facebook, Inc. and its affiliates.
+ *
+ * This source code is licensed under the MIT license found in the
+ * LICENSE file in the root directory of this source tree.
+ */
+/** @license React v17.0.2
+ * react-dom.production.min.js
+ *
+ * Copyright (c) Facebook, Inc. and its affiliates.
+ *
+ * This source code is licensed under the MIT license found in the
+ * LICENSE file in the root directory of this source tree.
+ */
+/** @license React v17.0.2
+ * react-jsx-runtime.production.min.js
+ *
+ * Copyright (c) Facebook, Inc. and its affiliates.
+ *
+ * This source code is licensed under the MIT license found in the
+ * LICENSE file in the root directory of this source tree.
+ */
+/** @license React v17.0.2
+ * react.production.min.js
+ *
+ * Copyright (c) Facebook, Inc. and its affiliates.
+ *
+ * This source code is licensed under the MIT license found in the
+ * LICENSE file in the root directory of this source tree.
+ */
+/**!
+ * FlexSearch.js v0.7.21 (Bundle)
+ * Copyright 2018-2021 Nextapps GmbH
+ * Author: Thomas Wilkerling
+ * Licence: Apache-2.0
+ * https://github.com/nextapps-de/flexsearch
+ */

app/build/static/js/main.ca662570.js ADDED Viewed

The diff for this file is too large to render. See raw diff

app/build/static/js/main.ca662570.js.LICENSE.txt ADDED Viewed

	@@ -0,0 +1,60 @@

+/*
+object-assign
+(c) Sindre Sorhus
+@license MIT
+*/
+/*! regenerator-runtime -- Copyright (c) 2014-present, Facebook, Inc. -- license (MIT): https://github.com/facebook/regenerator/blob/main/LICENSE */
+/**
+ * @license
+ * Lodash <https://lodash.com/>
+ * Copyright OpenJS Foundation and other contributors <https://openjsf.org/>
+ * Released under MIT license <https://lodash.com/license>
+ * Based on Underscore.js 1.8.3 <http://underscorejs.org/LICENSE>
+ * Copyright Jeremy Ashkenas, DocumentCloud and Investigative Reporters & Editors
+ */
+/** @license React v0.20.2
+ * scheduler.production.min.js
+ *
+ * Copyright (c) Facebook, Inc. and its affiliates.
+ *
+ * This source code is licensed under the MIT license found in the
+ * LICENSE file in the root directory of this source tree.
+ */
+/** @license React v17.0.2
+ * react-dom.production.min.js
+ *
+ * Copyright (c) Facebook, Inc. and its affiliates.
+ *
+ * This source code is licensed under the MIT license found in the
+ * LICENSE file in the root directory of this source tree.
+ */
+/** @license React v17.0.2
+ * react-jsx-runtime.production.min.js
+ *
+ * Copyright (c) Facebook, Inc. and its affiliates.
+ *
+ * This source code is licensed under the MIT license found in the
+ * LICENSE file in the root directory of this source tree.
+ */
+/** @license React v17.0.2
+ * react.production.min.js
+ *
+ * Copyright (c) Facebook, Inc. and its affiliates.
+ *
+ * This source code is licensed under the MIT license found in the
+ * LICENSE file in the root directory of this source tree.
+ */
+/**!
+ * FlexSearch.js v0.7.21 (Bundle)
+ * Copyright 2018-2021 Nextapps GmbH
+ * Author: Thomas Wilkerling
+ * Licence: Apache-2.0
+ * https://github.com/nextapps-de/flexsearch
+ */

app/build/static/js/main.ed69b879.js ADDED Viewed

The diff for this file is too large to render. See raw diff

app/build/static/js/main.ed69b879.js.LICENSE.txt ADDED Viewed

	@@ -0,0 +1,60 @@

+/*
+object-assign
+(c) Sindre Sorhus
+@license MIT
+*/
+/*! regenerator-runtime -- Copyright (c) 2014-present, Facebook, Inc. -- license (MIT): https://github.com/facebook/regenerator/blob/main/LICENSE */
+/**
+ * @license
+ * Lodash <https://lodash.com/>
+ * Copyright OpenJS Foundation and other contributors <https://openjsf.org/>
+ * Released under MIT license <https://lodash.com/license>
+ * Based on Underscore.js 1.8.3 <http://underscorejs.org/LICENSE>
+ * Copyright Jeremy Ashkenas, DocumentCloud and Investigative Reporters & Editors
+ */
+/** @license React v0.20.2
+ * scheduler.production.min.js
+ *
+ * Copyright (c) Facebook, Inc. and its affiliates.
+ *
+ * This source code is licensed under the MIT license found in the
+ * LICENSE file in the root directory of this source tree.
+ */
+/** @license React v17.0.2
+ * react-dom.production.min.js
+ *
+ * Copyright (c) Facebook, Inc. and its affiliates.
+ *
+ * This source code is licensed under the MIT license found in the
+ * LICENSE file in the root directory of this source tree.
+ */
+/** @license React v17.0.2
+ * react-jsx-runtime.production.min.js
+ *
+ * Copyright (c) Facebook, Inc. and its affiliates.
+ *
+ * This source code is licensed under the MIT license found in the
+ * LICENSE file in the root directory of this source tree.
+ */
+/** @license React v17.0.2
+ * react.production.min.js
+ *
+ * Copyright (c) Facebook, Inc. and its affiliates.
+ *
+ * This source code is licensed under the MIT license found in the
+ * LICENSE file in the root directory of this source tree.
+ */
+/**!
+ * FlexSearch.js v0.7.21 (Bundle)
+ * Copyright 2018-2021 Nextapps GmbH
+ * Author: Thomas Wilkerling
+ * Licence: Apache-2.0
+ * https://github.com/nextapps-de/flexsearch
+ */

app/build/static/media/WorkSans-Black.67c2c5a144333953880b.ttf ADDED Viewed

Binary file (192 kB). View file

app/build/static/media/WorkSans-Bold.2bea7a7f7d052c74da25.ttf ADDED Viewed

Binary file (193 kB). View file

app/build/static/media/WorkSans-Regular.bb287b894b27372d8ea7.ttf ADDED Viewed

Binary file (192 kB). View file

app/build/static/media/WorkSans-SemiBold.1e98db4eb705b586728e.ttf ADDED Viewed

Binary file (193 kB). View file

app/build/static/media/coffee-machine-lineal.ee32631219cc3986f861.gif ADDED Viewed

benchmark.py ADDED Viewed

	@@ -0,0 +1,108 @@

+# #!/usr/bin/env python3
+#
+# import argparse
+# import os
+# import time
+# import numpy as np
+# import nvidia_smi
+# import psutil
+# import torch
+#
+# from model_manager import ModelManager
+# from schema import Config, HDStrategy, SDSampler
+#
+# try:
+#     torch._C._jit_override_can_fuse_on_cpu(False)
+#     torch._C._jit_override_can_fuse_on_gpu(False)
+#     torch._C._jit_set_texpr_fuser_enabled(False)
+#     torch._C._jit_set_nvfuser_enabled(False)
+# except:
+#     pass
+#
+# NUM_THREADS = str(4)
+#
+# os.environ["OMP_NUM_THREADS"] = NUM_THREADS
+# os.environ["OPENBLAS_NUM_THREADS"] = NUM_THREADS
+# os.environ["MKL_NUM_THREADS"] = NUM_THREADS
+# os.environ["VECLIB_MAXIMUM_THREADS"] = NUM_THREADS
+# os.environ["NUMEXPR_NUM_THREADS"] = NUM_THREADS
+# if os.environ.get("CACHE_DIR"):
+#     os.environ["TORCH_HOME"] = os.environ["CACHE_DIR"]
+#
+#
+# def run_model(model, size):
+#     # RGB
+#     image = np.random.randint(0, 256, (size[0], size[1], 3)).astype(np.uint8)
+#     mask = np.random.randint(0, 255, size).astype(np.uint8)
+#
+#     config = Config(
+#         ldm_steps=2,
+#         hd_strategy=HDStrategy.ORIGINAL,
+#         hd_strategy_crop_margin=128,
+#         hd_strategy_crop_trigger_size=128,
+#         hd_strategy_resize_limit=128,
+#         prompt="a fox is sitting on a bench",
+#         sd_steps=5,
+#         sd_sampler=SDSampler.ddim
+#     )
+#     model(image, mask, config)
+#
+#
+# def benchmark(model, times: int, empty_cache: bool):
+#     sizes = [(512, 512)]
+#
+#     nvidia_smi.nvmlInit()
+#     device_id = 0
+#     handle = nvidia_smi.nvmlDeviceGetHandleByIndex(device_id)
+#
+#     def format(metrics):
+#         return f"{np.mean(metrics):.2f} ± {np.std(metrics):.2f}"
+#
+#     process = psutil.Process(os.getpid())
+#     # 每个 size 给出显存和内存占用的指标
+#     for size in sizes:
+#         torch.cuda.empty_cache()
+#         time_metrics = []
+#         cpu_metrics = []
+#         memory_metrics = []
+#         gpu_memory_metrics = []
+#         for _ in range(times):
+#             start = time.time()
+#             run_model(model, size)
+#             torch.cuda.synchronize()
+#
+#             # cpu_metrics.append(process.cpu_percent())
+#             time_metrics.append((time.time() - start) * 1000)
+#             memory_metrics.append(process.memory_info().rss / 1024 / 1024)
+#             gpu_memory_metrics.append(nvidia_smi.nvmlDeviceGetMemoryInfo(handle).used / 1024 / 1024)
+#
+#         print(f"size: {size}".center(80, "-"))
+#         # print(f"cpu: {format(cpu_metrics)}")
+#         print(f"latency: {format(time_metrics)}ms")
+#         print(f"memory: {format(memory_metrics)} MB")
+#         print(f"gpu memory: {format(gpu_memory_metrics)} MB")
+#
+#     nvidia_smi.nvmlShutdown()
+#
+#
+# def get_args_parser():
+#     parser = argparse.ArgumentParser()
+#     parser.add_argument("--name")
+#     parser.add_argument("--device", default="cuda", type=str)
+#     parser.add_argument("--times", default=10, type=int)
+#     parser.add_argument("--empty-cache", action="store_true")
+#     return parser.parse_args()
+#
+#
+# if __name__ == "__main__":
+#     args = get_args_parser()
+#     device = torch.device(args.device)
+#     model = ModelManager(
+#         name=args.name,
+#         device=device,
+#         sd_run_local=True,
+#         disable_nsfw=True,
+#         sd_cpu_textencoder=True,
+#         hf_access_token="123"
+#     )
+#     benchmark(model, args.times, args.empty_cache)

const.py ADDED Viewed

	@@ -0,0 +1,80 @@

+import os
+MPS_SUPPORT_MODELS = [
+    "instruct_pix2pix",
+    "sd1.5",
+    "anything4",
+    "realisticVision1.4",
+    "sd2",
+    "paint_by_example"
+]
+DEFAULT_MODEL = "lama"
+AVAILABLE_MODELS = [
+    "lama",
+    "ldm",
+    "zits",
+    "mat",
+    "fcf",
+    "sd1.5",
+    "anything4",
+    "realisticVision1.4",
+    "cv2",
+    "manga",
+    "sd2",
+    "paint_by_example",
+    "instruct_pix2pix",
+]
+AVAILABLE_DEVICES = ["cuda", "cpu", "mps"]
+DEFAULT_DEVICE = 'cuda'
+NO_HALF_HELP = """
+Using full precision model.
+If your generate result is always black or green, use this argument. (sd/paint_by_exmaple)
+"""
+CPU_OFFLOAD_HELP = """
+Offloads all models to CPU, significantly reducing vRAM usage. (sd/paint_by_example)
+"""
+DISABLE_NSFW_HELP = """
+Disable NSFW checker. (sd/paint_by_example)
+"""
+SD_CPU_TEXTENCODER_HELP = """
+Run Stable Diffusion text encoder model on CPU to save GPU memory.
+"""
+LOCAL_FILES_ONLY_HELP = """
+Use local files only, not connect to Hugging Face server. (sd/paint_by_example)
+"""
+ENABLE_XFORMERS_HELP = """
+Enable xFormers optimizations. Requires xformers package has been installed. See: https://github.com/facebookresearch/xformers (sd/paint_by_example)
+"""
+DEFAULT_MODEL_DIR = os.getenv(
+    "XDG_CACHE_HOME",
+    os.path.join(os.path.expanduser("~"), ".cache")
+)
+MODEL_DIR_HELP = """
+Model download directory (by setting XDG_CACHE_HOME environment variable), by default model downloaded to ~/.cache
+"""
+OUTPUT_DIR_HELP = """
+Result images will be saved to output directory automatically without confirmation.
+"""
+INPUT_HELP = """
+If input is image, it will be loaded by default.
+If input is directory, you can browse and select image in file manager.
+"""
+GUI_HELP = """
+Launch Lama Cleaner as desktop app
+"""
+NO_GUI_AUTO_CLOSE_HELP = """
+Prevent backend auto close after the GUI window closed.
+"""

ext/__init__.py ADDED Viewed

	@@ -0,0 +1 @@


1	+ from .image_watermark_handler import ImageWatermarkHandler

ext/__pycache__/__init__.cpython-38.pyc ADDED Viewed

Binary file (297 Bytes). View file

ext/__pycache__/__init__.cpython-39.pyc ADDED Viewed

Binary file (284 Bytes). View file

ext/__pycache__/image_watermark_handler.cpython-38.pyc ADDED Viewed

Binary file (2.58 kB). View file

ext/__pycache__/image_watermark_handler.cpython-39.pyc ADDED Viewed

Binary file (2.56 kB). View file

ext/image_watermark_handler.py ADDED Viewed

	@@ -0,0 +1,93 @@

+import io
+import os
+import shutil
+import cv2 as cv
+import numpy as np
+import requests
+from PIL import Image
+class ImageWatermarkHandler:
+    def __init__(self):
+        pass
+    def index_watermark(self, img_path, block_size, step_len, threshold, temp_output_path):
+        """
+        该代码首先读取图像，并定义了块的大小和阈值。然后，它遍历图像的每个块，并计算每个块的标准差。
+        如果标准差小于阈值，则认为当前块变化微小。然后将图片保存到本地（可选）并返回 list
+        :param img_path: img路径
+        :param watermark_size: 定义块的大小
+        :param step_len: 遍历的步长
+        :param threshold: 定义阈值
+        :param temp_output_path: sub_img临时保存路径
+        :return:
+        """
+        # read img
+        img = cv.imread(img_path, cv.IMREAD_GRAYSCALE)  # 把图像转成单通道的灰度图输出
+        # get width and height
+        height, width = img.shape
+        print("cur gray img width: %s,height: %s,block_size:%s" % (width, height, block_size))
+        block_num = int(height * width // block_size)
+        print("total split block num :  %s" % (block_num))
+        # remove last res dir
+        # 不保存图片就不用创建文件夹
+        # if (os.path.exists(temp_output_path)):
+        #     shutil.rmtree(temp_output_path)
+        # os.mkdir(temp_output_path)
+        # save pixel index  to memory and lcoal file
+        list = []
+        # foreach block
+        for i in range(0, height, step_len):
+            for j in range(0, width, step_len):
+                # get pixel value
+                block = img[i:i + block_size, j:j + block_size]
+                # print("cur idx [%s,%s], block : %s " %(i,j,block))
+                # calculate std_dev
+                std_dev = np.std(block)
+                # print(" cur std_dev :{} ,cur threshold : {} ".format(std_dev, threshold))  # 测试的像素区域,w:45-65--->com
+                # 如果标准差小于阈值，则认为当前块变化微小
+                if std_dev < threshold and std_dev > 0:
+                    # save memory
+                    dict = {}
+                    dict['w'] = j
+                    dict['h'] = i
+                    list.append(dict)
+                    # save local file
+                    f = temp_output_path + "{}-{}.png".format(j, i)
+                    print("save split img =====> w : %s ,h : %s ,cur std_dev : %s,cur threshold : %s ".format(j, i,
+                                                                                                              std_dev,
+                                                                                                              threshold))  # 测试的像素区域,w:45-65--->com
+                    # 可以不保存图片
+                    # cv.imwrite(f, block)
+        return list
+    def get_mask(self, img_path, list, block_size, mask_img_path):
+        """
+        获取mask
+        :param img_path:
+        :param list:
+        :param block_size:
+        :param mask_img_path:
+        :return:
+        """
+        img = cv.imread(img_path, cv.IMREAD_COLOR)
+        # black color
+        img[:] = 0
+        for item in list:
+            w = int(item.get("w"))
+            h = int(item.get("h"))
+            x1, y1 = w, h  # 左上角坐标
+            x2, y2 = w + block_size, h + block_size  # 右下角坐标
+            # white color
+            img[y1:y2, x1:x2] = 255, 255, 255
+            # print(img[y1,x1])
+        # save
+        cv.imwrite(mask_img_path, img)
+        return img

ext/request_info.txt ADDED Viewed

	@@ -0,0 +1,44 @@

+参数
+image: (binary) 源图像文件流
+mask: (binary) 掩码图像字节流
+ldm模型配置：
+ldmSteps: 25
+ldmSampler: plms
+zitsWireframe: true 线框
+hdStrategy: Crop ---> 支持的策略有：Crop、Origin、Resize。Crop masking area from the original image to do inpainting.对GPU友好
+hdStrategyCropMargin: 196
+hdStrategyCropTrigerSize: 800 ---> 会变
+hdStrategyResizeLimit: 2048
+prompt:
+negativePrompt:
+croperX: -206
+croperY: -222
+croperHeight: 512
+croperWidth: 512
+useCroper: false
+sdMaskBlur: 5
+sdStrength: 0.75
+sdSteps: 50
+sdGuidanceScale: 7.5
+sdSampler: pndm
+sdSeed: -1
+sdMatchHistograms: false
+sdScale: 1
+cv2Radius: 5
+cv2Flag: INPAINT_NS
+paintByExampleSteps: 50
+paintByExampleGuidanceScale: 7.5
+paintByExampleSeed: -1 ---> 会变
+paintByExampleMaskBlur: 5
+paintByExampleMatchHistograms: false
+p2pSteps: 50
+p2pImageGuidanceScale: 1.5
+p2pGuidanceScale: 7.5
+sizeLimit: 99

ext/test.py ADDED Viewed

	@@ -0,0 +1,341 @@

+import io
+import cv2
+import numpy as np
+import requests
+from PIL import Image
+"""
+1、测试lama-cleaner的inpaint api
+"""
+def test_inpaint_api():
+    """
+    参数为image、mask
+    :return:
+    """
+    # 加载原始图像并将其转换为灰度图像
+    img_path = '/resources/jeyoo-img/img.png'
+    mask_img_path = '/resources/jeyoo-img/img_mask.png'
+    clean_img_path = "/resources/jeyoo-img/img_clean.png"
+    img = cv2.imread(img_path)
+    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
+    cv2.imshow('gray', gray)
+    cv2.imwrite(mask_img_path,gray)
+    # image_file_object = cv2.imread(img_path)
+    # mask_file_object =  cv2.imread(mask_img_path)
+    r = requests.post('http://127.0.0.1:7860/inpaint',
+                      files={
+                          'image': open(img_path, 'rb'),
+                          'mask': open(mask_img_path, 'rb')},
+                      data={
+                          'ldmSteps': 25,
+                          'ldmSampler': "plms",
+                          'zitsWireframe': bool(True),
+                          'hdStrategy': "Crop",
+                          'hdStrategyCropMargin': 196,
+                          'hdStrategyCropTrigerSize': 800,
+                          'hdStrategyResizeLimit': 2048,
+                          'prompt': "",
+                          'negativePrompt': "",
+                          'croperX': 58,
+                          'croperY': -26,
+                          'croperHeight': 512,
+                          'croperWidth': 512,
+                          'useCroper': bool(False),
+                          'sdMaskBlur': 5,
+                          'sdStrength': 0.75,
+                          'sdSteps': 50,
+                          'sdGuidanceScale': 7.5,
+                          'sdSampler': "pndm",
+                          'sdSeed': -1,
+                          'sdMatchHistograms': bool(False),
+                          'sdScale': 1,
+                          'cv2Radius': 5,
+                          'cv2Flag': "INPAINT_NS",
+                          'paintByExampleSteps': 50,
+                          'paintByExampleGuidanceScale': 7.5,
+                          'paintByExampleSeed': -1,
+                          'paintByExampleMaskBlur': 5,
+                          'paintByExampleMatchHistograms': bool(False),
+                          'p2pSteps': 50,
+                          'p2pImageGuidanceScale': 1.5,
+                          'p2pGuidanceScale': 7.5,
+                          'sizeLimit': 628
+                      },
+                      headers={'x-api-key': 'xxxx'}
+                      )
+    if (r.ok):
+        # r.content contains the bytes of the returned image
+        print(r)
+        image_data = r.content
+        # 将图片数据转换为图像对象
+        image = Image.open(io.BytesIO(image_data))
+        # 将图像对象保存到本地文件
+        image.save(clean_img_path)
+    else:
+        r.raise_for_status()
+"""
+2、测试从从image根据 watermark 获取 mask img
+"""
+def test_get_mask_by_gradient():
+    """
+    在这个示例代码中，我们首先读取一张图片，并将其转换为灰度图像。
+    然后，我们使用Sobel算子计算图像的梯度，并计算像素变化的平均值。
+    接着，我们将像素变化小于平均值一半的像素设置为0，得到一个二值掩码。
+    最后，我们使用findContours函数找到掩码中的轮廓，并使用boundingRect函数获取每个轮廓的矩形框。
+    最后，我们在原图上绘制矩形框，并显示结果。
+    """
+    # 读取图片
+    img = cv2.imread('../lama_cleaner_source_code/resources/jeyoo-img/image31.png')
+    # 转换为灰度图像
+    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
+    # 计算梯度
+    # src:gray图像
+    # ddepth:int类型的ddepth，输出图像的深度，若src.depth() = CV_32F, 取ddepth =-1/CV_32F/CV_64F
+    # dx、dy:X、y方向的差分阶数
+    grad_x = cv2.Sobel(gray, cv2.CV_32F, 1, 0)
+    grad_y = cv2.Sobel(gray, cv2.CV_32F, 0, 1)
+    # print("grad_x:{},grad_y:{}".format(grad_x,grad_y))
+    # gray_x、gray_y是矩阵，第二个参数是权重
+    grad = cv2.addWeighted(grad_x, 0.5, grad_y, 0.5, 0)
+    # 计算像素变化小的区域
+    # 对数组中的每一个元素求其绝对值。
+    grad_abs = np.absolute(grad)
+    # 相似变化的平均值
+    grad_mean = np.mean(grad_abs)
+    # 0.7是最大
+    grad_threshold = grad_mean * 0.5
+    grad_mask = grad_abs < grad_threshold
+    # for i, element in enumerate(grad_abs):
+    #     print("第 {} 行 ".format(i))
+    #     for j in enumerate(element):
+    #         print("{}".format(j))
+    print("grad_mean:{},grad_threshold:{}".format(grad_mean,grad_threshold)) # grad_mean:93.4161148071289,grad_threshold:46.70805740356445
+    # 获取像素变化小的区域的矩形框
+    contours, hierarchy = cv2.findContours(grad_mask.astype(np.uint8), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
+    # print(contours)
+    rects = [cv2.boundingRect(cnt) for cnt in contours]
+    # 在原图上绘制矩形框
+    for rect in rects:
+        cv2.rectangle(img, (rect[0], rect[1]), (rect[0] + rect[2], rect[1] + rect[3]), (0, 0, 255), 2)
+    # 显示结果
+    cv2.imshow('image', img)
+    cv2.waitKey(0)
+    cv2.destroyAllWindows()
+    cv2.imwrite('../lama_cleaner_source_code/resources/jeyoo-img/out/imgage31.png',img)
+def test_get_mask_by_watermark():
+    """
+    从image根据 watermark_img 获取mask img
+    要从图像中获取水印位置的掩码图像，您可以使用OpenCV中的模板匹配技术。模板匹配是一种在图像中查找给定模板的技术，它可以用于检测图像中的水印位置。
+    以下是使用OpenCV从图像中获取水印位置的掩码图像的步骤：
+    加载原始图像和水印图像。
+    将水印图像转换为灰度图像。
+    使用OpenCV中的模板匹配函数（cv2.matchTemplate）在原始图像中查找水印图像的位置。
+    根据匹配结果创建掩码图像。在掩码图像中，将匹配位置设置为白色，其他位置设置为黑色。
+    :return:
+    """
+    # 加载原始图像和水印图像
+    img = cv2.imread('original_image.jpg')
+    watermark = cv2.imread('watermark_image.jpg')
+    # 将水印图像转换为灰度图像
+    watermark_gray = cv2.cvtColor(watermark, cv2.COLOR_BGR2GRAY)
+    # 使用模板匹配在原始图像中查找水印图像的位置
+    result = cv2.matchTemplate(img, watermark_gray, cv2.TM_CCOEFF_NORMED)
+    # 根据匹配结果创建掩码图像
+    threshold = 0.8
+    mask = np.zeros_like(result)
+    mask[result >= threshold] = 255
+    # 显示掩码图像
+    cv2.imshow('Mask', mask)
+    # 等待用户按下任意键
+    cv2.waitKey(0)
+    # 释放窗口
+    cv2.destroyAllWindows()
+def test_get_mask_by_watermark2():
+    """
+    在上面的代码中，我们使用Canny算子对灰度图像进行边缘检测，并使用cv2.findContours函数对边缘图像进行轮廓检测。
+    然后，我们对每个轮廓进行形状分析，并根据筛选出的轮廓创建掩码图像。
+    在筛选轮廓时，我们使用了cv2.isContourConvex函数来排除非凸形状的轮廓。您可以根据需要调整形状分析的参数来获得更好的结果。
+    :return:
+    """
+    # 加载原始图像并将其转换为灰度图像
+    img = cv2.imread('../lama_cleaner_source_code/resources/jeyoo-img/img.png')
+    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
+    cv2.imshow('gray', gray)
+    cv2.waitKey(0)
+    # 对灰度图像进行边缘检测
+    edges = cv2.Canny(gray, 50, 150)
+    cv2.imshow('edges', edges)
+    cv2.waitKey(0)
+    # 对边缘图像进行轮廓检测
+    contours, hierarchy = cv2.findContours(edges, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
+    # 对每个轮廓进行形状分析，筛选出可能是水印的轮廓
+    watermark_contours = []
+    for contour in contours:
+        perimeter = cv2.arcLength(contour, True)
+        approx = cv2.approxPolyDP(contour, 0.02 * perimeter, True)
+        print(len(approx))
+        if len(approx) == 6 and cv2.isContourConvex(approx):
+            watermark_contours.append(approx)
+    # 根据筛选出的轮廓，创建掩码图像
+    mask = np.zeros_like(gray)
+    for contour in watermark_contours:
+        cv2.drawContours(mask, [contour], 0, 255, -1)
+    # 显示掩码图像
+    cv2.imshow('Mask', mask)
+    # 等待用户按下任意键
+    cv2.waitKey(0)
+    # 释放窗口
+    cv2.destroyAllWindows()
+def test_get_mask_img_watermark3():
+    """
+    要为水印区域生成掩码图像，您可以使用OpenCV中的矩形掩码。以下是生成掩码图像的步骤：
+    创建一个与原始图像大小相同的黑色图像，作为掩码图像。
+    使用OpenCV中的 cv2.rectangle() 函数在掩码图像上绘制一个矩形，该矩形覆盖水印区域。
+    将掩码图像转换为灰度图像，并使用OpenCV中的 cv2.threshold() 函数将其二值化，以便将水印区域设置为白色，其余区域设置为黑色。
+    :return:
+    """
+    img = cv2.imread('../lama_cleaner_source_code/resources/jeyoo-img/img.png')
+    # 创建掩码图像
+    mask = np.zeros_like(img)
+    # 定义水印区域
+    x, y, w, h = 3, 18, 29, 29
+    # 在掩码图像上绘制矩形
+    cv2.rectangle(mask, (x, y), (x + w, y + h), (255, 255, 255), -1)
+    # 将掩码图像转换为灰度图像
+    mask_gray = cv2.cvtColor(mask, cv2.COLOR_BGR2GRAY)
+    # 将掩码图像二值化
+    _, mask_binary = cv2.threshold(mask_gray, 1, 255, cv2.THRESH_BINARY)
+    # 显示掩码图像
+    cv2.imshow('Mask', mask_binary)
+    cv2.waitKey(0)
+    cv2.destroyAllWindows()
+"""
+3、测试从image根据 watermark 获取 mask img
+"""
+def test_get_mask_by_inrange():
+    """
+    方式1
+    :return:
+    """
+    img_path = '../lama_cleaner_docker/resources/jeyoo2-shuiyin.png'
+    img = cv2.imread(img_path)
+    # 转换为灰度图像
+    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
+    # 计算掩码,删除像素位于该区间的
+    mask = cv2.inRange(gray, 240, 255)
+    cv2.imshow('mask1', mask)
+    cv2.waitKey(0)
+    """
+    方式2
+    """
+    # 读取图片
+    img = cv2.imread(img_path)
+    cv2.imshow("img" ,img)
+    cv2.waitKey(0)
+    # 将图片转换为HSV颜色空间
+    hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
+    cv2.imshow("hsv", hsv)
+    cv2.waitKey(0)
+    means, dev = cv2.meanStdDev(img)
+    print("means:{}".format(means))
+    # means: [[227.75111119]
+    #         [228.73636804]
+    #         [225.89541678]]
+    # 定义颜色范围
+    lower_color = np.array([100, 100, 100])
+    upper_color = np.array([255, 255, 255])
+    # 创建掩码
+    mask2 = cv2.inRange(hsv, lower_color, upper_color)
+    cv2.imshow("mask2", mask2)
+    cv2.waitKey(0)
+    # 获取选定区域
+    result = cv2.bitwise_and(img, img, mask=mask)
+    # 显示结果
+    cv2.imshow('image', result)
+    cv2.waitKey(0)
+    cv2.destroyAllWindows()
+"""
+测试opencv的inpaint修复方法
+"""
+def test_repire_old_img_by_cv():
+    img = cv2.imread('../lama_cleaner_source_code/resources/jeyoo-shuiyin.png')
+    mask = cv2.imread('../lama_cleaner_source_code/resources/jeyoo2-shuiyin_mask.jpg', cv2.IMREAD_GRAYSCALE)
+    dst = cv2.inpaint(img, mask, 3, cv2.INPAINT_TELEA)
+    cv2.imshow('dst', dst)
+    cv2.waitKey(0)
+    cv2.destroyAllWindows()

file_manager/__init__.py ADDED Viewed

	@@ -0,0 +1 @@


1	+ from .file_manager import FileManager

file_manager/__pycache__/__init__.cpython-38.pyc ADDED Viewed

Binary file (285 Bytes). View file

file_manager/__pycache__/__init__.cpython-39.pyc ADDED Viewed

Binary file (272 Bytes). View file

file_manager/__pycache__/file_manager.cpython-38.pyc ADDED Viewed

Binary file (7.32 kB). View file

file_manager/__pycache__/file_manager.cpython-39.pyc ADDED Viewed

Binary file (7.29 kB). View file

file_manager/__pycache__/storage_backends.cpython-38.pyc ADDED Viewed

Binary file (2.07 kB). View file

file_manager/__pycache__/storage_backends.cpython-39.pyc ADDED Viewed

Binary file (2.09 kB). View file

file_manager/__pycache__/utils.cpython-38.pyc ADDED Viewed

Binary file (1.72 kB). View file

file_manager/__pycache__/utils.cpython-39.pyc ADDED Viewed

Binary file (1.7 kB). View file

file_manager/file_manager.py ADDED Viewed

	@@ -0,0 +1,264 @@

+# Copy from https://github.com/silentsokolov/flask-thumbnails/blob/master/flask_thumbnails/thumbnail.py
+import os
+from datetime import datetime
+import cv2
+import time
+from io import BytesIO
+from pathlib import Path
+import numpy as np
+from watchdog.events import FileSystemEventHandler
+from watchdog.observers import Observer
+from PIL import Image, ImageOps, PngImagePlugin
+from loguru import logger
+LARGE_ENOUGH_NUMBER = 100
+PngImagePlugin.MAX_TEXT_CHUNK = LARGE_ENOUGH_NUMBER * (1024**2)
+from .storage_backends import FilesystemStorageBackend
+from .utils import aspect_to_string, generate_filename, glob_img
+class FileManager(FileSystemEventHandler):
+    def __init__(self, app=None):
+        self.app = app
+        self._default_root_directory = "media"
+        self._default_thumbnail_directory = "media"
+        self._default_root_url = "/"
+        self._default_thumbnail_root_url = "/"
+        self._default_format = "JPEG"
+        self.output_dir: Path = None
+        if app is not None:
+            self.init_app(app)
+        self.image_dir_filenames = []
+        self.output_dir_filenames = []
+        self.image_dir_observer = None
+        self.output_dir_observer = None
+        self.modified_time = {
+            "image": datetime.utcnow(),
+            "output": datetime.utcnow(),
+        }
+    def start(self):
+        self.image_dir_filenames = self._media_names(self.root_directory)
+        self.output_dir_filenames = self._media_names(self.output_dir)
+        logger.info(f"Start watching image directory: {self.root_directory}")
+        self.image_dir_observer = Observer()
+        self.image_dir_observer.schedule(self, self.root_directory, recursive=False)
+        self.image_dir_observer.start()
+        logger.info(f"Start watching output directory: {self.output_dir}")
+        self.output_dir_observer = Observer()
+        self.output_dir_observer.schedule(self, self.output_dir, recursive=False)
+        self.output_dir_observer.start()
+    def on_modified(self, event):
+        if not os.path.isdir(event.src_path):
+            return
+        if event.src_path == str(self.root_directory):
+            logger.info(f"Image directory {event.src_path} modified")
+            self.image_dir_filenames = self._media_names(self.root_directory)
+            self.modified_time["image"] = datetime.utcnow()
+        elif event.src_path == str(self.output_dir):
+            logger.info(f"Output directory {event.src_path} modified")
+            self.output_dir_filenames = self._media_names(self.output_dir)
+            self.modified_time["output"] = datetime.utcnow()
+    def init_app(self, app):
+        if self.app is None:
+            self.app = app
+        app.thumbnail_instance = self
+        if not hasattr(app, "extensions"):
+            app.extensions = {}
+        if "thumbnail" in app.extensions:
+            raise RuntimeError("Flask-thumbnail extension already initialized")
+        app.extensions["thumbnail"] = self
+        app.config.setdefault("THUMBNAIL_MEDIA_ROOT", self._default_root_directory)
+        app.config.setdefault(
+            "THUMBNAIL_MEDIA_THUMBNAIL_ROOT", self._default_thumbnail_directory
+        )
+        app.config.setdefault("THUMBNAIL_MEDIA_URL", self._default_root_url)
+        app.config.setdefault(
+            "THUMBNAIL_MEDIA_THUMBNAIL_URL", self._default_thumbnail_root_url
+        )
+        app.config.setdefault("THUMBNAIL_DEFAULT_FORMAT", self._default_format)
+    @property
+    def root_directory(self):
+        path = self.app.config["THUMBNAIL_MEDIA_ROOT"]
+        if os.path.isabs(path):
+            return path
+        else:
+            return os.path.join(self.app.root_path, path)
+    @property
+    def thumbnail_directory(self):
+        path = self.app.config["THUMBNAIL_MEDIA_THUMBNAIL_ROOT"]
+        if os.path.isabs(path):
+            return path
+        else:
+            return os.path.join(self.app.root_path, path)
+    @property
+    def root_url(self):
+        return self.app.config["THUMBNAIL_MEDIA_URL"]
+    @property
+    def media_names(self):
+        # return self.image_dir_filenames
+        return self._media_names(self.root_directory)
+    @property
+    def output_media_names(self):
+        return self._media_names(self.output_dir)
+        # return self.output_dir_filenames
+    @staticmethod
+    def _media_names(directory: Path):
+        names = sorted([it.name for it in glob_img(directory)])
+        res = []
+        for name in names:
+            path = os.path.join(directory, name)
+            img = Image.open(path)
+            res.append(
+                {
+                    "name": name,
+                    "height": img.height,
+                    "width": img.width,
+                    "ctime": os.path.getctime(path),
+                }
+            )
+        return res
+    @property
+    def thumbnail_url(self):
+        return self.app.config["THUMBNAIL_MEDIA_THUMBNAIL_URL"]
+    def get_thumbnail(
+        self, directory: Path, original_filename: str, width, height, **options
+    ):
+        storage = FilesystemStorageBackend(self.app)
+        crop = options.get("crop", "fit")
+        background = options.get("background")
+        quality = options.get("quality", 90)
+        original_path, original_filename = os.path.split(original_filename)
+        original_filepath = os.path.join(directory, original_path, original_filename)
+        image = Image.open(BytesIO(storage.read(original_filepath)))
+        # keep ratio resize
+        if width is not None:
+            height = int(image.height * width / image.width)
+        else:
+            width = int(image.width * height / image.height)
+        thumbnail_size = (width, height)
+        thumbnail_filename = generate_filename(
+            original_filename,
+            aspect_to_string(thumbnail_size),
+            crop,
+            background,
+            quality,
+        )
+        thumbnail_filepath = os.path.join(
+            self.thumbnail_directory, original_path, thumbnail_filename
+        )
+        thumbnail_url = os.path.join(
+            self.thumbnail_url, original_path, thumbnail_filename
+        )
+        if storage.exists(thumbnail_filepath):
+            return thumbnail_url, (width, height)
+        try:
+            image.load()
+        except (IOError, OSError):
+            self.app.logger.warning("Thumbnail not load image: %s", original_filepath)
+            return thumbnail_url, (width, height)
+        # get original image format
+        options["format"] = options.get("format", image.format)
+        image = self._create_thumbnail(
+            image, thumbnail_size, crop, background=background
+        )
+        raw_data = self.get_raw_data(image, **options)
+        storage.save(thumbnail_filepath, raw_data)
+        return thumbnail_url, (width, height)
+    def get_raw_data(self, image, **options):
+        data = {
+            "format": self._get_format(image, **options),
+            "quality": options.get("quality", 90),
+        }
+        _file = BytesIO()
+        image.save(_file, **data)
+        return _file.getvalue()
+    @staticmethod
+    def colormode(image, colormode="RGB"):
+        if colormode == "RGB" or colormode == "RGBA":
+            if image.mode == "RGBA":
+                return image
+            if image.mode == "LA":
+                return image.convert("RGBA")
+            return image.convert(colormode)
+        if colormode == "GRAY":
+            return image.convert("L")
+        return image.convert(colormode)
+    @staticmethod
+    def background(original_image, color=0xFF):
+        size = (max(original_image.size),) * 2
+        image = Image.new("L", size, color)
+        image.paste(
+            original_image,
+            tuple(map(lambda x: (x[0] - x[1]) / 2, zip(size, original_image.size))),
+        )
+        return image
+    def _get_format(self, image, **options):
+        if options.get("format"):
+            return options.get("format")
+        if image.format:
+            return image.format
+        return self.app.config["THUMBNAIL_DEFAULT_FORMAT"]
+    def _create_thumbnail(self, image, size, crop="fit", background=None):
+        try:
+            resample = Image.Resampling.LANCZOS
+        except AttributeError:  # pylint: disable=raise-missing-from
+            resample = Image.ANTIALIAS
+        if crop == "fit":
+            image = ImageOps.fit(image, size, resample)
+        else:
+            image = image.copy()
+            image.thumbnail(size, resample=resample)
+        if background is not None:
+            image = self.background(image)
+        image = self.colormode(image)
+        return image

file_manager/storage_backends.py ADDED Viewed

	@@ -0,0 +1,46 @@

+# Copy from https://github.com/silentsokolov/flask-thumbnails/blob/master/flask_thumbnails/storage_backends.py
+import errno
+import os
+from abc import ABC, abstractmethod
+class BaseStorageBackend(ABC):
+    def __init__(self, app=None):
+        self.app = app
+    @abstractmethod
+    def read(self, filepath, mode="rb", **kwargs):
+        raise NotImplementedError
+    @abstractmethod
+    def exists(self, filepath):
+        raise NotImplementedError
+    @abstractmethod
+    def save(self, filepath, data):
+        raise NotImplementedError
+class FilesystemStorageBackend(BaseStorageBackend):
+    def read(self, filepath, mode="rb", **kwargs):
+        with open(filepath, mode) as f:  # pylint: disable=unspecified-encoding
+            return f.read()
+    def exists(self, filepath):
+        return os.path.exists(filepath)
+    def save(self, filepath, data):
+        directory = os.path.dirname(filepath)
+        if not os.path.exists(directory):
+            try:
+                os.makedirs(directory)
+            except OSError as e:
+                if e.errno != errno.EEXIST:
+                    raise
+        if not os.path.isdir(directory):
+            raise IOError("{} is not a directory".format(directory))
+        with open(filepath, "wb") as f:
+            f.write(data)

file_manager/utils.py ADDED Viewed

	@@ -0,0 +1,67 @@

+# Copy from: https://github.com/silentsokolov/flask-thumbnails/blob/master/flask_thumbnails/utils.py
+import importlib
+import os
+from pathlib import Path
+from typing import Union
+def generate_filename(original_filename, *options):
+    name, ext = os.path.splitext(original_filename)
+    for v in options:
+        if v:
+            name += "_%s" % v
+    name += ext
+    return name
+def parse_size(size):
+    if isinstance(size, int):
+        # If the size parameter is a single number, assume square aspect.
+        return [size, size]
+    if isinstance(size, (tuple, list)):
+        if len(size) == 1:
+            # If single value tuple/list is provided, exand it to two elements
+            return size + type(size)(size)
+        return size
+    try:
+        thumbnail_size = [int(x) for x in size.lower().split("x", 1)]
+    except ValueError:
+        raise ValueError(  # pylint: disable=raise-missing-from
+            "Bad thumbnail size format. Valid format is INTxINT."
+        )
+    if len(thumbnail_size) == 1:
+        # If the size parameter only contains a single integer, assume square aspect.
+        thumbnail_size.append(thumbnail_size[0])
+    return thumbnail_size
+def aspect_to_string(size):
+    if isinstance(size, str):
+        return size
+    return "x".join(map(str, size))
+IMG_SUFFIX = {'.jpg', '.jpeg', '.png', '.JPG', '.JPEG', '.PNG'}
+def glob_img(p: Union[Path, str], recursive: bool = False):
+    p = Path(p)
+    if p.is_file() and p.suffix in IMG_SUFFIX:
+        yield p
+    else:
+        if recursive:
+            files = Path(p).glob("**/*.*")
+        else:
+            files = Path(p).glob("*.*")
+        for it in files:
+            if it.suffix not in IMG_SUFFIX:
+                continue
+            yield it

helper.py ADDED Viewed

	@@ -0,0 +1,284 @@

+import io
+import os
+import sys
+from typing import List, Optional
+from urllib.parse import urlparse
+import cv2
+from PIL import Image, ImageOps
+import numpy as np
+import torch
+from const import MPS_SUPPORT_MODELS
+from loguru import logger
+from torch.hub import download_url_to_file, get_dir
+import hashlib
+def md5sum(filename):
+    md5 = hashlib.md5()
+    with open(filename, "rb") as f:
+        for chunk in iter(lambda: f.read(128 * md5.block_size), b""):
+            md5.update(chunk)
+    return md5.hexdigest()
+def switch_mps_device(model_name, device):
+    if model_name not in MPS_SUPPORT_MODELS and str(device) == "mps":
+        logger.info(f"{model_name} not support mps, switch to cpu")
+        return torch.device("cpu")
+    return device
+def get_cache_path_by_url(url):
+    parts = urlparse(url)
+    hub_dir = get_dir()
+    model_dir = os.path.join(hub_dir, "checkpoints")
+    if not os.path.isdir(model_dir):
+        os.makedirs(model_dir)
+    filename = os.path.basename(parts.path)
+    cached_file = os.path.join(model_dir, filename)
+    return cached_file
+def download_model(url, model_md5: str = None):
+    cached_file = get_cache_path_by_url(url)
+    if not os.path.exists(cached_file):
+        sys.stderr.write('Downloading: "{}" to {}\n'.format(url, cached_file))
+        hash_prefix = None
+        download_url_to_file(url, cached_file, hash_prefix, progress=True)
+        if model_md5:
+            _md5 = md5sum(cached_file)
+            if model_md5 == _md5:
+                logger.info(f"Download model success, md5: {_md5}")
+            else:
+                try:
+                    os.remove(cached_file)
+                    logger.error(
+                        f"Model md5: {_md5}, expected md5: {model_md5}, wrong model deleted. Please restart lama-cleaner."
+                        f"If you still have errors, please try download model manually first https://lama-cleaner-docs.vercel.app/install/download_model_manually.\n"
+                    )
+                except:
+                    logger.error(
+                        f"Model md5: {_md5}, expected md5: {model_md5}, please delete {cached_file} and restart lama-cleaner."
+                    )
+                exit(-1)
+    return cached_file
+def ceil_modulo(x, mod):
+    if x % mod == 0:
+        return x
+    return (x // mod + 1) * mod
+def handle_error(model_path, model_md5, e):
+    _md5 = md5sum(model_path)
+    if _md5 != model_md5:
+        try:
+            os.remove(model_path)
+            logger.error(
+                f"Model md5: {_md5}, expected md5: {model_md5}, wrong model deleted. Please restart lama-cleaner."
+                f"If you still have errors, please try download model manually first https://lama-cleaner-docs.vercel.app/install/download_model_manually.\n"
+            )
+        except:
+            logger.error(
+                f"Model md5: {_md5}, expected md5: {model_md5}, please delete {model_path} and restart lama-cleaner."
+            )
+    else:
+        logger.error(
+            f"Failed to load model {model_path},"
+            f"please submit an issue at https://github.com/Sanster/lama-cleaner/issues and include a screenshot of the error:\n{e}"
+        )
+    exit(-1)
+def load_jit_model(url_or_path, device, model_md5: str):
+    if os.path.exists(url_or_path):
+        model_path = url_or_path
+    else:
+        model_path = download_model(url_or_path, model_md5)
+    logger.info(f"Loading model from: {model_path}")
+    try:
+        model = torch.jit.load(model_path, map_location="cpu").to(device)
+    except Exception as e:
+        handle_error(model_path, model_md5, e)
+    model.eval()
+    return model
+def load_model(model: torch.nn.Module, url_or_path, device, model_md5):
+    if os.path.exists(url_or_path):
+        model_path = url_or_path
+    else:
+        model_path = download_model(url_or_path, model_md5)
+    try:
+        logger.info(f"Loading model from: {model_path}")
+        state_dict = torch.load(model_path, map_location="cpu")
+        model.load_state_dict(state_dict, strict=True)
+        model.to(device)
+    except Exception as e:
+        handle_error(model_path, model_md5, e)
+    model.eval()
+    return model
+def numpy_to_bytes(image_numpy: np.ndarray, ext: str) -> bytes:
+    data = cv2.imencode(
+        f".{ext}",
+        image_numpy,
+        [int(cv2.IMWRITE_JPEG_QUALITY), 100, int(cv2.IMWRITE_PNG_COMPRESSION), 0],
+    )[1]
+    image_bytes = data.tobytes()
+    return image_bytes
+def pil_to_bytes(pil_img, ext: str, exif=None) -> bytes:
+    with io.BytesIO() as output:
+        pil_img.save(output, format=ext, exif=exif, quality=95)
+        image_bytes = output.getvalue()
+    return image_bytes
+def load_img(img_bytes, gray: bool = False, return_exif: bool = False):
+    alpha_channel = None
+    image = Image.open(io.BytesIO(img_bytes))
+    try:
+        if return_exif:
+            exif = image.getexif()
+    except:
+        exif = None
+        logger.error("Failed to extract exif from image")
+    try:
+        image = ImageOps.exif_transpose(image)
+    except:
+        pass
+    if gray:
+        image = image.convert("L")
+        np_img = np.array(image)
+    else:
+        if image.mode == "RGBA":
+            np_img = np.array(image)
+            alpha_channel = np_img[:, :, -1]
+            np_img = cv2.cvtColor(np_img, cv2.COLOR_RGBA2RGB)
+        else:
+            image = image.convert("RGB")
+            np_img = np.array(image)
+    if return_exif:
+        return np_img, alpha_channel, exif
+    return np_img, alpha_channel
+def norm_img(np_img):
+    if len(np_img.shape) == 2:
+        np_img = np_img[:, :, np.newaxis]
+    np_img = np.transpose(np_img, (2, 0, 1))
+    np_img = np_img.astype("float32") / 255
+    return np_img
+def resize_max_size(
+    np_img, size_limit: int, interpolation=cv2.INTER_CUBIC
+) -> np.ndarray:
+    # Resize image's longer size to size_limit if longer size larger than size_limit
+    h, w = np_img.shape[:2]
+    if max(h, w) > size_limit:
+        ratio = size_limit / max(h, w)
+        new_w = int(w * ratio + 0.5)
+        new_h = int(h * ratio + 0.5)
+        return cv2.resize(np_img, dsize=(new_w, new_h), interpolation=interpolation)
+    else:
+        return np_img
+def pad_img_to_modulo(
+    img: np.ndarray, mod: int, square: bool = False, min_size: Optional[int] = None
+):
+    """
+    Args:
+        img: [H, W, C]
+        mod:
+        square: 是否为正方形
+        min_size:
+    Returns:
+    """
+    if len(img.shape) == 2:
+        img = img[:, :, np.newaxis]
+    height, width = img.shape[:2]
+    out_height = ceil_modulo(height, mod)
+    out_width = ceil_modulo(width, mod)
+    if min_size is not None:
+        assert min_size % mod == 0
+        out_width = max(min_size, out_width)
+        out_height = max(min_size, out_height)
+    if square:
+        max_size = max(out_height, out_width)
+        out_height = max_size
+        out_width = max_size
+    return np.pad(
+        img,
+        ((0, out_height - height), (0, out_width - width), (0, 0)),
+        mode="symmetric",
+    )
+def boxes_from_mask(mask: np.ndarray) -> List[np.ndarray]:
+    """
+    Args:
+        mask: (h, w, 1)  0~255
+    Returns:
+    """
+    height, width = mask.shape[:2]
+    _, thresh = cv2.threshold(mask, 127, 255, 0)
+    contours, _ = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
+    boxes = []
+    for cnt in contours:
+        x, y, w, h = cv2.boundingRect(cnt)
+        box = np.array([x, y, x + w, y + h]).astype(int)
+        box[::2] = np.clip(box[::2], 0, width)
+        box[1::2] = np.clip(box[1::2], 0, height)
+        boxes.append(box)
+    return boxes
+def only_keep_largest_contour(mask: np.ndarray) -> List[np.ndarray]:
+    """
+    Args:
+        mask: (h, w)  0~255
+    Returns:
+    """
+    _, thresh = cv2.threshold(mask, 127, 255, 0)
+    contours, _ = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
+    max_area = 0
+    max_index = -1
+    for i, cnt in enumerate(contours):
+        area = cv2.contourArea(cnt)
+        if area > max_area:
+            max_area = area
+            max_index = i
+    if max_index != -1:
+        new_mask = np.zeros_like(mask)
+        return cv2.drawContours(new_mask, contours, max_index, 255, -1)
+    else:
+        return mask

interactive_seg.py ADDED Viewed

	@@ -0,0 +1,203 @@

+import os
+import cv2
+from typing import Tuple, List
+import torch
+import torch.nn.functional as F
+from loguru import logger
+from pydantic import BaseModel
+import numpy as np
+from helper import only_keep_largest_contour, load_jit_model
+class Click(BaseModel):
+    # [y, x]
+    coords: Tuple[float, float]
+    is_positive: bool
+    indx: int
+    @property
+    def coords_and_indx(self):
+        return (*self.coords, self.indx)
+    def scale(self, x_ratio: float, y_ratio: float) -> 'Click':
+        return Click(
+            coords=(self.coords[0] * x_ratio, self.coords[1] * y_ratio),
+            is_positive=self.is_positive,
+            indx=self.indx
+        )
+class ResizeTrans:
+    def __init__(self, size=480):
+        super().__init__()
+        self.crop_height = size
+        self.crop_width = size
+    def transform(self, image_nd, clicks_lists):
+        assert image_nd.shape[0] == 1 and len(clicks_lists) == 1
+        image_height, image_width = image_nd.shape[2:4]
+        self.image_height = image_height
+        self.image_width = image_width
+        image_nd_r = F.interpolate(image_nd, (self.crop_height, self.crop_width), mode='bilinear', align_corners=True)
+        y_ratio = self.crop_height / image_height
+        x_ratio = self.crop_width / image_width
+        clicks_lists_resized = []
+        for clicks_list in clicks_lists:
+            clicks_list_resized = [click.scale(y_ratio, x_ratio) for click in clicks_list]
+            clicks_lists_resized.append(clicks_list_resized)
+        return image_nd_r, clicks_lists_resized
+    def inv_transform(self, prob_map):
+        new_prob_map = F.interpolate(prob_map, (self.image_height, self.image_width), mode='bilinear',
+                                     align_corners=True)
+        return new_prob_map
+class ISPredictor(object):
+    def __init__(
+        self,
+        model,
+        device,
+        open_kernel_size: int,
+        dilate_kernel_size: int,
+        net_clicks_limit=None,
+        zoom_in=None,
+        infer_size=384,
+    ):
+        self.model = model
+        self.open_kernel_size = open_kernel_size
+        self.dilate_kernel_size = dilate_kernel_size
+        self.net_clicks_limit = net_clicks_limit
+        self.device = device
+        self.zoom_in = zoom_in
+        self.infer_size = infer_size
+        # self.transforms = [zoom_in] if zoom_in is not None else []
+    def __call__(self, input_image: torch.Tensor, clicks: List[Click], prev_mask):
+        """
+        Args:
+            input_image: [1, 3, H, W]  [0~1]
+            clicks: List[Click]
+            prev_mask: [1, 1, H, W]
+        Returns:
+        """
+        transforms = [ResizeTrans(self.infer_size)]
+        input_image = torch.cat((input_image, prev_mask), dim=1)
+        # image_nd resized to infer_size
+        for t in transforms:
+            image_nd, clicks_lists = t.transform(input_image, [clicks])
+        # image_nd.shape = [1, 4, 256, 256]
+        # points_nd.sha[e = [1, 2, 3]
+        # clicks_lists[0][0] Click 类
+        points_nd = self.get_points_nd(clicks_lists)
+        pred_logits = self.model(image_nd, points_nd)
+        pred = torch.sigmoid(pred_logits)
+        pred = self.post_process(pred)
+        prediction = F.interpolate(pred, mode='bilinear', align_corners=True,
+                                   size=image_nd.size()[2:])
+        for t in reversed(transforms):
+            prediction = t.inv_transform(prediction)
+        # if self.zoom_in is not None and self.zoom_in.check_possible_recalculation():
+        #    return self.get_prediction(clicker)
+        return prediction.cpu().numpy()[0, 0]
+    def post_process(self, pred: torch.Tensor) -> torch.Tensor:
+        pred_mask = pred.cpu().numpy()[0][0]
+        # morph_open to remove small noise
+        kernel_size = self.open_kernel_size
+        kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (kernel_size, kernel_size))
+        pred_mask = cv2.morphologyEx(pred_mask, cv2.MORPH_OPEN, kernel, iterations=1)
+        # Why dilate: make region slightly larger to avoid missing some pixels, this generally works better
+        dilate_kernel_size = self.dilate_kernel_size
+        if dilate_kernel_size > 1:
+            kernel = cv2.getStructuringElement(cv2.MORPH_DILATE, (dilate_kernel_size, dilate_kernel_size))
+            pred_mask = cv2.dilate(pred_mask, kernel, 1)
+        return torch.from_numpy(pred_mask).unsqueeze(0).unsqueeze(0)
+    def get_points_nd(self, clicks_lists):
+        total_clicks = []
+        num_pos_clicks = [sum(x.is_positive for x in clicks_list) for clicks_list in clicks_lists]
+        num_neg_clicks = [len(clicks_list) - num_pos for clicks_list, num_pos in zip(clicks_lists, num_pos_clicks)]
+        num_max_points = max(num_pos_clicks + num_neg_clicks)
+        if self.net_clicks_limit is not None:
+            num_max_points = min(self.net_clicks_limit, num_max_points)
+        num_max_points = max(1, num_max_points)
+        for clicks_list in clicks_lists:
+            clicks_list = clicks_list[:self.net_clicks_limit]
+            pos_clicks = [click.coords_and_indx for click in clicks_list if click.is_positive]
+            pos_clicks = pos_clicks + (num_max_points - len(pos_clicks)) * [(-1, -1, -1)]
+            neg_clicks = [click.coords_and_indx for click in clicks_list if not click.is_positive]
+            neg_clicks = neg_clicks + (num_max_points - len(neg_clicks)) * [(-1, -1, -1)]
+            total_clicks.append(pos_clicks + neg_clicks)
+        return torch.tensor(total_clicks, device=self.device)
+INTERACTIVE_SEG_MODEL_URL = os.environ.get(
+    "INTERACTIVE_SEG_MODEL_URL",
+    "https://github.com/Sanster/models/releases/download/clickseg_pplnet/clickseg_pplnet.pt",
+)
+INTERACTIVE_SEG_MODEL_MD5 = os.environ.get("INTERACTIVE_SEG_MODEL_MD5", "8ca44b6e02bca78f62ec26a3c32376cf")
+class InteractiveSeg:
+    def __init__(self, infer_size=384, open_kernel_size=3, dilate_kernel_size=3):
+        device = torch.device('cpu')
+        model = load_jit_model(INTERACTIVE_SEG_MODEL_URL, device, INTERACTIVE_SEG_MODEL_MD5).eval()
+        self.predictor = ISPredictor(model, device,
+                                     infer_size=infer_size,
+                                     open_kernel_size=open_kernel_size,
+                                     dilate_kernel_size=dilate_kernel_size)
+    def __call__(self, image, clicks, prev_mask=None):
+        """
+        Args:
+            image: [H,W,C] RGB
+            clicks:
+        Returns:
+        """
+        image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
+        image = torch.from_numpy((image / 255).transpose(2, 0, 1)).unsqueeze(0).float()
+        if prev_mask is None:
+            mask = torch.zeros_like(image[:, :1, :, :])
+        else:
+            logger.info('InteractiveSeg run with prev_mask')
+            mask = torch.from_numpy(prev_mask / 255).unsqueeze(0).unsqueeze(0).float()
+        pred_probs = self.predictor(image, clicks, mask)
+        pred_mask = pred_probs > 0.5
+        pred_mask = (pred_mask * 255).astype(np.uint8)
+        # Find largest contour
+        # pred_mask = only_keep_largest_contour(pred_mask)
+        # To simplify frontend process, add mask brush color here
+        fg = pred_mask == 255
+        bg = pred_mask != 255
+        pred_mask = cv2.cvtColor(pred_mask, cv2.COLOR_GRAY2BGRA)
+        # frontend brush color "ffcc00bb"
+        pred_mask[bg] = 0
+        pred_mask[fg] = [255, 203, 0, int(255 * 0.73)]
+        pred_mask = cv2.cvtColor(pred_mask, cv2.COLOR_BGRA2RGBA)
+        return pred_mask

make_gif.py ADDED Viewed

	@@ -0,0 +1,125 @@

+import io
+import math
+from pathlib import Path
+from PIL import Image, ImageDraw
+def keep_ratio_resize(img, size, resample=Image.BILINEAR):
+    if img.width > img.height:
+        w = size
+        h = int(img.height * size / img.width)
+    else:
+        h = size
+        w = int(img.width * size / img.height)
+    return img.resize((w, h), resample)
+def cubic_bezier(p1, p2, duration: int, frames: int):
+    """
+    Args:
+        p1:
+        p2:
+        duration: Total duration of the curve
+        frames:
+    Returns:
+    """
+    x0, y0 = (0, 0)
+    x1, y1 = p1
+    x2, y2 = p2
+    x3, y3 = (1, 1)
+    def cal_y(t):
+        return math.pow(1 - t, 3) * y0 + \
+               3 * math.pow(1 - t, 2) * t * y1 + \
+               3 * (1 - t) * math.pow(t, 2) * y2 + \
+               math.pow(t, 3) * y3
+    def cal_x(t):
+        return math.pow(1 - t, 3) * x0 + \
+               3 * math.pow(1 - t, 2) * t * x1 + \
+               3 * (1 - t) * math.pow(t, 2) * x2 + \
+               math.pow(t, 3) * x3
+    res = []
+    for t in range(0, 1 * frames, duration):
+        t = t / frames
+        res.append((cal_x(t), cal_y(t)))
+    res.append((1, 0))
+    return res
+def make_compare_gif(
+    clean_img: Image.Image,
+    src_img: Image.Image,
+    max_side_length: int = 600,
+    splitter_width: int = 5,
+    splitter_color=(255, 203, 0, int(255 * 0.73))
+):
+    if clean_img.size != src_img.size:
+        clean_img = clean_img.resize(src_img.size, Image.BILINEAR)
+    duration_per_frame = 20
+    num_frames = 50
+    # erase-in-out
+    cubic_bezier_points = cubic_bezier((0.33, 0), (0.66, 1), 1, num_frames)
+    cubic_bezier_points.reverse()
+    max_side_length = min(max_side_length, max(clean_img.size))
+    src_img = keep_ratio_resize(src_img, max_side_length)
+    clean_img = keep_ratio_resize(clean_img, max_side_length)
+    width, height = src_img.size
+    # Generate images to make Gif from right to left
+    images = []
+    for i in range(num_frames):
+        new_frame = Image.new('RGB', (width, height))
+        new_frame.paste(clean_img, (0, 0))
+        left = int(cubic_bezier_points[i][0] * width)
+        cropped_src_img = src_img.crop((left, 0, width, height))
+        new_frame.paste(cropped_src_img, (left, 0, width, height))
+        if i != num_frames - 1:
+            # draw a yellow splitter on the edge of the cropped image
+            draw = ImageDraw.Draw(new_frame)
+            draw.line([(left, 0), (left, height)], width=splitter_width, fill=splitter_color)
+        images.append(new_frame)
+    for i in range(10):
+        images.append(src_img)
+    cubic_bezier_points.reverse()
+    # Generate images to make Gif from left to right
+    for i in range(num_frames):
+        new_frame = Image.new('RGB', (width, height))
+        new_frame.paste(src_img, (0, 0))
+        right = int(cubic_bezier_points[i][0] * width)
+        cropped_src_img = clean_img.crop((0, 0, right, height))
+        new_frame.paste(cropped_src_img, (0, 0, right, height))
+        if i != num_frames - 1:
+            # draw a yellow splitter on the edge of the cropped image
+            draw = ImageDraw.Draw(new_frame)
+            draw.line([(right, 0), (right, height)], width=splitter_width, fill=splitter_color)
+        images.append(new_frame)
+    images.append(clean_img)
+    img_byte_arr = io.BytesIO()
+    clean_img.save(
+        img_byte_arr,
+        format='GIF',
+        save_all=True,
+        include_color_table=True,
+        append_images=images,
+        optimize=False,
+        duration=duration_per_frame,
+        loop=0
+    )
+    return img_byte_arr.getvalue()

model/__init__.py ADDED Viewed

File without changes

model/__pycache__/__init__.cpython-38.pyc ADDED Viewed

Binary file (160 Bytes). View file

model/__pycache__/__init__.cpython-39.pyc ADDED Viewed

Binary file (160 Bytes). View file

model/__pycache__/base.cpython-38.pyc ADDED Viewed

Binary file (7.82 kB). View file

model/__pycache__/base.cpython-39.pyc ADDED Viewed

Binary file (7.8 kB). View file

model/__pycache__/fcf.cpython-38.pyc ADDED Viewed

Binary file (34.2 kB). View file

model/__pycache__/instruct_pix2pix.cpython-38.pyc ADDED Viewed

Binary file (2.78 kB). View file

model/__pycache__/lama.cpython-38.pyc ADDED Viewed

Binary file (1.84 kB). View file