Upload 11 files

.gitignore

@@ -0,0 +1,21 @@
+# Python byte code, etc.
+__pycache__/
+
+# C/C++ object files/libraries
+*.o
+*.so
+
+# macOS
+**/.DS_Store
+
+# tmp
+~*
+
+# pretrained_models
+pretrained_models/big-lama
+# pytracking/pretrain/
+*.pth
+
+results/*
+
+

LICENSE

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

app.py

@@ -0,0 +1,462 @@
+import os
+import sys
+
+sys.path.append(os.path.abspath(os.path.dirname(os.getcwd())))
+# os.chdir("../")
+import cv2
+import gradio as gr
+import numpy as np
+from pathlib import Path
+from matplotlib import pyplot as plt
+import torch
+import tempfile
+
+from stable_diffusion_inpaint import fill_img_with_sd, replace_img_with_sd
+from lama_inpaint import (
+    inpaint_img_with_lama,
+    build_lama_model,
+    inpaint_img_with_builded_lama,
+)
+from utils import (
+    load_img_to_array,
+    save_array_to_img,
+    dilate_mask,
+    show_mask,
+    show_points,
+)
+from PIL import Image
+from segment_anything import SamPredictor, sam_model_registry
+import argparse
+
+
+def setup_args(parser):
+    parser.add_argument(
+        "--lama_config",
+        type=str,
+        default="./lama/configs/prediction/default.yaml",
+        help="The path to the config file of lama model. "
+        "Default: the config of big-lama",
+    )
+    parser.add_argument(
+        "--lama_ckpt",
+        type=str,
+        default="pretrained_models/big-lama",
+        help="The path to the lama checkpoint.",
+    )
+    parser.add_argument(
+        "--sam_ckpt",
+        type=str,
+        default="./pretrained_models/sam_vit_h_4b8939.pth",
+        help="The path to the SAM checkpoint to use for mask generation.",
+    )
+
+
+def mkstemp(suffix, dir=None):
+    fd, path = tempfile.mkstemp(suffix=f"{suffix}", dir=dir)
+    os.close(fd)
+    return Path(path)
+
+
+def get_sam_feat(img):
+    model["sam"].set_image(img)
+    features = model["sam"].features
+    orig_h = model["sam"].orig_h
+    orig_w = model["sam"].orig_w
+    input_h = model["sam"].input_h
+    input_w = model["sam"].input_w
+    model["sam"].reset_image()
+    return features, orig_h, orig_w, input_h, input_w
+
+
+def get_fill_img_with_sd(image, mask, image_resolution, text_prompt):
+    device = "cuda" if torch.cuda.is_available() else "cpu"
+    if len(mask.shape) == 3:
+        mask = mask[:, :, 0]
+    np_image = np.array(image, dtype=np.uint8)
+    H, W, C = np_image.shape
+    np_image = HWC3(np_image)
+    np_image = resize_image(np_image, image_resolution)
+    mask = cv2.resize(
+        mask, (np_image.shape[1], np_image.shape[0]), interpolation=cv2.INTER_NEAREST
+    )
+
+    img_fill = fill_img_with_sd(np_image, mask, text_prompt, device=device)
+    img_fill = img_fill.astype(np.uint8)
+    return img_fill
+
+
+def get_replace_img_with_sd(image, mask, image_resolution, text_prompt):
+    device = "cuda" if torch.cuda.is_available() else "cpu"
+    if len(mask.shape) == 3:
+        mask = mask[:, :, 0]
+    np_image = np.array(image, dtype=np.uint8)
+    H, W, C = np_image.shape
+    np_image = HWC3(np_image)
+    np_image = resize_image(np_image, image_resolution)
+    mask = cv2.resize(
+        mask, (np_image.shape[1], np_image.shape[0]), interpolation=cv2.INTER_NEAREST
+    )
+
+    img_replaced = replace_img_with_sd(np_image, mask, text_prompt, device=device)
+    img_replaced = img_replaced.astype(np.uint8)
+    return img_replaced
+
+
+def HWC3(x):
+    assert x.dtype == np.uint8
+    if x.ndim == 2:
+        x = x[:, :, None]
+    assert x.ndim == 3
+    H, W, C = x.shape
+    assert C == 1 or C == 3 or C == 4
+    if C == 3:
+        return x
+    if C == 1:
+        return np.concatenate([x, x, x], axis=2)
+    if C == 4:
+        color = x[:, :, 0:3].astype(np.float32)
+        alpha = x[:, :, 3:4].astype(np.float32) / 255.0
+        y = color * alpha + 255.0 * (1.0 - alpha)
+        y = y.clip(0, 255).astype(np.uint8)
+        return y
+
+
+def resize_image(input_image, resolution):
+    H, W, C = input_image.shape
+    k = float(resolution) / min(H, W)
+    H = int(np.round(H * k / 64.0)) * 64
+    W = int(np.round(W * k / 64.0)) * 64
+    img = cv2.resize(
+        input_image,
+        (W, H),
+        interpolation=cv2.INTER_LANCZOS4 if k > 1 else cv2.INTER_AREA,
+    )
+    return img
+
+
+def resize_points(clicked_points, original_shape, resolution):
+    original_height, original_width, _ = original_shape
+    original_height = float(original_height)
+    original_width = float(original_width)
+
+    scale_factor = float(resolution) / min(original_height, original_width)
+    resized_points = []
+
+    for point in clicked_points:
+        x, y, lab = point
+        resized_x = int(round(x * scale_factor))
+        resized_y = int(round(y * scale_factor))
+        resized_point = (resized_x, resized_y, lab)
+        resized_points.append(resized_point)
+
+    return resized_points
+
+
+def get_click_mask(
+    clicked_points, features, orig_h, orig_w, input_h, input_w, dilate_kernel_size
+):
+    # model['sam'].set_image(image)
+    model["sam"].is_image_set = True
+    model["sam"].features = features
+    model["sam"].orig_h = orig_h
+    model["sam"].orig_w = orig_w
+    model["sam"].input_h = input_h
+    model["sam"].input_w = input_w
+
+    # Separate the points and labels
+    points, labels = zip(*[(point[:2], point[2]) for point in clicked_points])
+
+    # Convert the points and labels to numpy arrays
+    input_point = np.array(points)
+    input_label = np.array(labels)
+
+    masks, _, _ = model["sam"].predict(
+        point_coords=input_point,
+        point_labels=input_label,
+        multimask_output=False,
+    )
+    if dilate_kernel_size is not None:
+        masks = [dilate_mask(mask, dilate_kernel_size) for mask in masks]
+    else:
+        masks = [mask for mask in masks]
+
+    return masks
+
+
+def process_image_click(
+    original_image,
+    point_prompt,
+    clicked_points,
+    image_resolution,
+    features,
+    orig_h,
+    orig_w,
+    input_h,
+    input_w,
+    dilate_kernel_size,
+    evt: gr.SelectData,
+):
+    if clicked_points is None:
+        clicked_points = []
+
+    # print("Received click event:", evt)
+    if original_image is None:
+        # print("No image loaded.")
+        return None, clicked_points, None
+
+    clicked_coords = evt.index
+    if clicked_coords is None:
+        # print("No valid coordinates received.")
+        return None, clicked_points, None
+
+    x, y = clicked_coords
+    label = point_prompt
+    lab = 1 if label == "Foreground Point" else 0
+    clicked_points.append((x, y, lab))
+    # print("Updated points list:", clicked_points)
+
+    input_image = np.array(original_image, dtype=np.uint8)
+    H, W, C = input_image.shape
+    input_image = HWC3(input_image)
+    img = resize_image(input_image, image_resolution)
+    # print("Processed image size:", img.shape)
+
+    resized_points = resize_points(clicked_points, input_image.shape, image_resolution)
+    mask_click_np = get_click_mask(
+        resized_points, features, orig_h, orig_w, input_h, input_w, dilate_kernel_size
+    )
+    mask_click_np = np.transpose(mask_click_np, (1, 2, 0)) * 255.0
+    mask_image = HWC3(mask_click_np.astype(np.uint8))
+    mask_image = cv2.resize(mask_image, (W, H), interpolation=cv2.INTER_LINEAR)
+    # print("Mask image prepared.")
+
+    edited_image = input_image
+    for x, y, lab in clicked_points:
+        color = (255, 0, 0) if lab == 1 else (0, 0, 255)
+        edited_image = cv2.circle(edited_image, (x, y), 20, color, -1)
+
+    opacity_mask = 0.75
+    opacity_edited = 1.0
+    overlay_image = cv2.addWeighted(
+        edited_image,
+        opacity_edited,
+        (mask_image * np.array([0 / 255, 255 / 255, 0 / 255])).astype(np.uint8),
+        opacity_mask,
+        0,
+    )
+
+    no_mask_overlay = edited_image.copy()
+
+    return no_mask_overlay, overlay_image, clicked_points, mask_image
+
+
+def image_upload(image, image_resolution):
+    if image is None:
+        return None, None, None, None, None, None
+    else:
+        np_image = np.array(image, dtype=np.uint8)
+        H, W, C = np_image.shape
+        np_image = HWC3(np_image)
+        np_image = resize_image(np_image, image_resolution)
+        features, orig_h, orig_w, input_h, input_w = get_sam_feat(np_image)
+        return image, features, orig_h, orig_w, input_h, input_w
+
+
+def get_inpainted_img(image, mask, image_resolution):
+    lama_config = args.lama_config
+    device = "cuda" if torch.cuda.is_available() else "cpu"
+    if len(mask.shape) == 3:
+        mask = mask[:, :, 0]
+    img_inpainted = inpaint_img_with_builded_lama(
+        model["lama"], image, mask, lama_config, device=device
+    )
+    return img_inpainted
+
+
+# get args
+parser = argparse.ArgumentParser()
+setup_args(parser)
+args = parser.parse_args(sys.argv[1:])
+# build models
+model = {}
+# build the sam model
+model_type = "vit_h"
+ckpt_p = args.sam_ckpt
+model_sam = sam_model_registry[model_type](checkpoint=ckpt_p)
+device = "cuda" if torch.cuda.is_available() else "cpu"
+model_sam.to(device=device)
+model["sam"] = SamPredictor(model_sam)
+
+# build the lama model
+lama_config = args.lama_config
+lama_ckpt = args.lama_ckpt
+device = "cuda" if torch.cuda.is_available() else "cpu"
+model["lama"] = build_lama_model(lama_config, lama_ckpt, device=device)
+
+button_size = (100, 50)
+with gr.Blocks() as demo:
+    clicked_points = gr.State([])
+    # origin_image = gr.State(None)
+    click_mask = gr.State(None)
+    features = gr.State(None)
+    orig_h = gr.State(None)
+    orig_w = gr.State(None)
+    input_h = gr.State(None)
+    input_w = gr.State(None)
+
+    with gr.Row():
+        with gr.Column(variant="panel"):
+            with gr.Row():
+                gr.Markdown("## Upload an image and click the region you want to edit.")
+            with gr.Row():
+                source_image_click = gr.Image(
+                    type="numpy",
+                    interactive=True,
+                    label="Upload and Edit Image",
+                )
+
+                image_edit_complete = gr.Image(
+                    type="numpy",
+                    interactive=False,
+                    label="Editing Complete",
+                )
+            with gr.Row():
+                point_prompt = gr.Radio(
+                    choices=["Foreground Point", "Background Point"],
+                    value="Foreground Point",
+                    label="Point Label",
+                    interactive=True,
+                    show_label=False,
+                )
+                image_resolution = gr.Slider(
+                    label="Image Resolution",
+                    minimum=256,
+                    maximum=768,
+                    value=512,
+                    step=64,
+                )
+                dilate_kernel_size = gr.Slider(
+                    label="Dilate Kernel Size", minimum=0, maximum=30, value=15, step=1
+                )
+        with gr.Column(variant="panel"):
+            with gr.Row():
+                gr.Markdown("## Control Panel")
+            text_prompt = gr.Textbox(label="Text Prompt")
+            lama = gr.Button("Inpaint Image", variant="primary")
+            fill_sd = gr.Button("Fill Anything with SD", variant="primary")
+            replace_sd = gr.Button("Replace Anything with SD", variant="primary")
+            clear_button_image = gr.Button(value="Reset", variant="secondary")
+
+    # todo: maybe we can delete this row, for it's unnecessary to show the original mask for customers
+    with gr.Row(variant="panel"):
+        with gr.Column():
+            with gr.Row():
+                gr.Markdown("## Mask")
+            with gr.Row():
+                click_mask = gr.Image(
+                    type="numpy",
+                    label="Click Mask",
+                    interactive=False,
+                )
+        with gr.Column():
+            with gr.Row():
+                gr.Markdown("## Image Removed with Mask")
+            with gr.Row():
+                img_rm_with_mask = gr.Image(
+                    type="numpy",
+                    label="Image Removed with Mask",
+                    interactive=False,
+                )
+
+        with gr.Column():
+            with gr.Row():
+                gr.Markdown("## Fill Anything with Mask")
+            with gr.Row():
+                img_fill_with_mask = gr.Image(
+                    type="numpy",
+                    label="Image Fill Anything with Mask",
+                    interactive=False,
+                )
+
+        with gr.Column():
+            with gr.Row():
+                gr.Markdown("## Replace Anything with Mask")
+            with gr.Row():
+                img_replace_with_mask = gr.Image(
+                    type="numpy",
+                    label="Image Replace Anything with Mask",
+                    interactive=False,
+                )
+
+    source_image_click.upload(
+        image_upload,
+        inputs=[source_image_click, image_resolution],
+        outputs=[source_image_click, features, orig_h, orig_w, input_h, input_w],
+    )
+
+    source_image_click.select(
+        process_image_click,
+        inputs=[
+            source_image_click,
+            point_prompt,
+            clicked_points,
+            image_resolution,
+            features,
+            orig_h,
+            orig_w,
+            input_h,
+            input_w,
+            dilate_kernel_size,
+        ],
+        outputs=[source_image_click, image_edit_complete, clicked_points, click_mask],
+        show_progress=True,
+        queue=True,
+    )
+
+    lama.click(
+        get_inpainted_img,
+        inputs=[source_image_click, click_mask, image_resolution],
+        outputs=[img_rm_with_mask],
+    )
+
+    fill_sd.click(
+        get_fill_img_with_sd,
+        inputs=[source_image_click, click_mask, image_resolution, text_prompt],
+        outputs=[img_fill_with_mask],
+    )
+
+    replace_sd.click(
+        get_replace_img_with_sd,
+        inputs=[source_image_click, click_mask, image_resolution, text_prompt],
+        outputs=[img_replace_with_mask],
+    )
+
+    def reset(*args):
+        return [None for _ in args]
+
+    clear_button_image.click(
+        reset,
+        inputs=[
+            source_image_click,
+            image_edit_complete,
+            clicked_points,
+            click_mask,
+            features,
+            img_rm_with_mask,
+            img_fill_with_mask,
+            img_replace_with_mask,
+        ],
+        outputs=[
+            source_image_click,
+            image_edit_complete,
+            clicked_points,
+            click_mask,
+            features,
+            img_rm_with_mask,
+            img_fill_with_mask,
+            img_replace_with_mask,
+        ],
+    )
+
+if __name__ == "__main__":
+    demo.launch(debug=False, show_error=True)

fill_anything.py

@@ -0,0 +1,137 @@
+import cv2
+import sys
+import argparse
+import numpy as np
+import torch
+from pathlib import Path
+from matplotlib import pyplot as plt
+from typing import Any, Dict, List
+
+from sam_segment import predict_masks_with_sam
+from stable_diffusion_inpaint import fill_img_with_sd
+from utils import load_img_to_array, save_array_to_img, dilate_mask, \
+    show_mask, show_points, get_clicked_point
+
+
+def setup_args(parser):
+    parser.add_argument(
+        "--input_img", type=str, required=True,
+        help="Path to a single input img",
+    )
+    parser.add_argument(
+        "--coords_type", type=str, required=True,
+        default="key_in", choices=["click", "key_in"], 
+        help="The way to select coords",
+    )
+    parser.add_argument(
+        "--point_coords", type=float, nargs='+', required=True,
+        help="The coordinate of the point prompt, [coord_W coord_H].",
+    )
+    parser.add_argument(
+        "--point_labels", type=int, nargs='+', required=True,
+        help="The labels of the point prompt, 1 or 0.",
+    )
+    parser.add_argument(
+        "--text_prompt", type=str, required=True,
+        help="Text prompt",
+    )
+    parser.add_argument(
+        "--dilate_kernel_size", type=int, default=None,
+        help="Dilate kernel size. Default: None",
+    )
+    parser.add_argument(
+        "--output_dir", type=str, required=True,
+        help="Output path to the directory with results.",
+    )
+    parser.add_argument(
+        "--sam_model_type", type=str,
+        default="vit_h", choices=['vit_h', 'vit_l', 'vit_b', 'vit_t'],
+        help="The type of sam model to load. Default: 'vit_h"
+    )
+    parser.add_argument(
+        "--sam_ckpt", type=str, required=True,
+        help="The path to the SAM checkpoint to use for mask generation.",
+    )
+    parser.add_argument(
+        "--seed", type=int,
+        help="Specify seed for reproducibility.",
+    )
+    parser.add_argument(
+        "--deterministic", action="store_true",
+        help="Use deterministic algorithms for reproducibility.",
+    )
+
+
+if __name__ == "__main__":
+    """Example usage:
+    python fill_anything.py \
+        --input_img FA_demo/FA1_dog.png \
+        --coords_type key_in \
+        --point_coords 750 500 \
+        --point_labels 1 \
+        --text_prompt "a teddy bear on a bench" \
+        --dilate_kernel_size 15 \
+        --output_dir ./results \
+        --sam_model_type "vit_h" \
+        --sam_ckpt sam_vit_h_4b8939.pth 
+    """
+    parser = argparse.ArgumentParser()
+    setup_args(parser)
+    args = parser.parse_args(sys.argv[1:])
+    device = "cuda" if torch.cuda.is_available() else "cpu"
+
+    if args.coords_type == "click":
+        latest_coords = get_clicked_point(args.input_img)
+    elif args.coords_type == "key_in":
+        latest_coords = args.point_coords
+    img = load_img_to_array(args.input_img)
+
+    masks, _, _ = predict_masks_with_sam(
+        img,
+        [latest_coords],
+        args.point_labels,
+        model_type=args.sam_model_type,
+        ckpt_p=args.sam_ckpt,
+        device=device,
+    )
+    masks = masks.astype(np.uint8) * 255
+
+    # dilate mask to avoid unmasked edge effect
+    if args.dilate_kernel_size is not None:
+        masks = [dilate_mask(mask, args.dilate_kernel_size) for mask in masks]
+
+    # visualize the segmentation results
+    img_stem = Path(args.input_img).stem
+    out_dir = Path(args.output_dir) / img_stem
+    out_dir.mkdir(parents=True, exist_ok=True)
+    for idx, mask in enumerate(masks):
+        # path to the results
+        mask_p = out_dir / f"mask_{idx}.png"
+        img_points_p = out_dir / f"with_points.png"
+        img_mask_p = out_dir / f"with_{Path(mask_p).name}"
+
+        # save the mask
+        save_array_to_img(mask, mask_p)
+
+        # save the pointed and masked image
+        dpi = plt.rcParams['figure.dpi']
+        height, width = img.shape[:2]
+        plt.figure(figsize=(width/dpi/0.77, height/dpi/0.77))
+        plt.imshow(img)
+        plt.axis('off')
+        show_points(plt.gca(), [latest_coords], args.point_labels,
+                    size=(width*0.04)**2)
+        plt.savefig(img_points_p, bbox_inches='tight', pad_inches=0)
+        show_mask(plt.gca(), mask, random_color=False)
+        plt.savefig(img_mask_p, bbox_inches='tight', pad_inches=0)
+        plt.close()
+
+    # fill the masked image
+    for idx, mask in enumerate(masks):
+        if args.seed is not None:
+            torch.manual_seed(args.seed)
+        mask_p = out_dir / f"mask_{idx}.png"
+        img_filled_p = out_dir / f"filled_with_{Path(mask_p).name}"
+        img_filled = fill_img_with_sd(
+            img, mask, args.text_prompt, device=device)
+        save_array_to_img(img_filled, img_filled_p)

lama_inpaint.py

@@ -0,0 +1,200 @@
+import os
+import sys
+import numpy as np
+import torch
+import yaml
+import glob
+import argparse
+from PIL import Image
+from omegaconf import OmegaConf
+from pathlib import Path
+
+os.environ['OMP_NUM_THREADS'] = '1'
+os.environ['OPENBLAS_NUM_THREADS'] = '1'
+os.environ['MKL_NUM_THREADS'] = '1'
+os.environ['VECLIB_MAXIMUM_THREADS'] = '1'
+os.environ['NUMEXPR_NUM_THREADS'] = '1'
+
+sys.path.insert(0, str(Path(__file__).resolve().parent / "lama"))
+from saicinpainting.evaluation.utils import move_to_device
+from saicinpainting.training.trainers import load_checkpoint
+from saicinpainting.evaluation.data import pad_tensor_to_modulo
+
+from utils import load_img_to_array, save_array_to_img
+
+
+@torch.no_grad()
+def inpaint_img_with_lama(
+        img: np.ndarray,
+        mask: np.ndarray,
+        config_p: str,
+        ckpt_p: str,
+        mod=8,
+        device="cuda"
+):
+    assert len(mask.shape) == 2
+    if np.max(mask) == 1:
+        mask = mask * 255
+    img = torch.from_numpy(img).float().div(255.)
+    mask = torch.from_numpy(mask).float()
+    predict_config = OmegaConf.load(config_p)
+    predict_config.model.path = ckpt_p
+    # device = torch.device(predict_config.device)
+    device = torch.device(device)
+
+    train_config_path = os.path.join(
+        predict_config.model.path, 'config.yaml')
+
+    with open(train_config_path, 'r') as f:
+        train_config = OmegaConf.create(yaml.safe_load(f))
+
+    train_config.training_model.predict_only = True
+    train_config.visualizer.kind = 'noop'
+
+    checkpoint_path = os.path.join(
+        predict_config.model.path, 'models',
+        predict_config.model.checkpoint
+    )
+    model = load_checkpoint(
+        train_config, checkpoint_path, strict=False, map_location='cpu')
+    model.freeze()
+    if not predict_config.get('refine', False):
+        model.to(device)
+
+    batch = {}
+    batch['image'] = img.permute(2, 0, 1).unsqueeze(0)
+    batch['mask'] = mask[None, None]
+    unpad_to_size = [batch['image'].shape[2], batch['image'].shape[3]]
+    batch['image'] = pad_tensor_to_modulo(batch['image'], mod)
+    batch['mask'] = pad_tensor_to_modulo(batch['mask'], mod)
+    batch = move_to_device(batch, device)
+    batch['mask'] = (batch['mask'] > 0) * 1
+
+    batch = model(batch)
+    cur_res = batch[predict_config.out_key][0].permute(1, 2, 0)
+    cur_res = cur_res.detach().cpu().numpy()
+
+    if unpad_to_size is not None:
+        orig_height, orig_width = unpad_to_size
+        cur_res = cur_res[:orig_height, :orig_width]
+
+    cur_res = np.clip(cur_res * 255, 0, 255).astype('uint8')
+    return cur_res
+
+
+def build_lama_model(        
+        config_p: str,
+        ckpt_p: str,
+        device="cuda"
+):
+    predict_config = OmegaConf.load(config_p)
+    predict_config.model.path = ckpt_p
+    device = torch.device(device)
+
+    train_config_path = os.path.join(
+        predict_config.model.path, 'config.yaml')
+
+    with open(train_config_path, 'r') as f:
+        train_config = OmegaConf.create(yaml.safe_load(f))
+
+    train_config.training_model.predict_only = True
+    train_config.visualizer.kind = 'noop'
+
+    checkpoint_path = os.path.join(
+        predict_config.model.path, 'models',
+        predict_config.model.checkpoint
+    )
+    model = load_checkpoint(train_config, checkpoint_path, strict=False)
+    model.to(device)
+    model.freeze()
+    return model
+
+
+@torch.no_grad()
+def inpaint_img_with_builded_lama(
+        model,
+        img: np.ndarray,
+        mask: np.ndarray,
+        config_p=None,
+        mod=8,
+        device="cuda"
+):
+    assert len(mask.shape) == 2
+    if np.max(mask) == 1:
+        mask = mask * 255
+    img = torch.from_numpy(img).float().div(255.)
+    mask = torch.from_numpy(mask).float()
+
+    batch = {}
+    batch['image'] = img.permute(2, 0, 1).unsqueeze(0)
+    batch['mask'] = mask[None, None]
+    unpad_to_size = [batch['image'].shape[2], batch['image'].shape[3]]
+    batch['image'] = pad_tensor_to_modulo(batch['image'], mod)
+    batch['mask'] = pad_tensor_to_modulo(batch['mask'], mod)
+    batch = move_to_device(batch, device)
+    batch['mask'] = (batch['mask'] > 0) * 1
+
+    batch = model(batch)
+    cur_res = batch["inpainted"][0].permute(1, 2, 0)
+    cur_res = cur_res.detach().cpu().numpy()
+
+    if unpad_to_size is not None:
+        orig_height, orig_width = unpad_to_size
+        cur_res = cur_res[:orig_height, :orig_width]
+
+    cur_res = np.clip(cur_res * 255, 0, 255).astype('uint8')
+    return cur_res
+
+
+
+def setup_args(parser):
+    parser.add_argument(
+        "--input_img", type=str, required=True,
+        help="Path to a single input img",
+    )
+    parser.add_argument(
+        "--input_mask_glob", type=str, required=True,
+        help="Glob to input masks",
+    )
+    parser.add_argument(
+        "--output_dir", type=str, required=True,
+        help="Output path to the directory with results.",
+    )
+    parser.add_argument(
+        "--lama_config", type=str,
+        default="./lama/configs/prediction/default.yaml",
+        help="The path to the config file of lama model. "
+             "Default: the config of big-lama",
+    )
+    parser.add_argument(
+        "--lama_ckpt", type=str, required=True,
+        help="The path to the lama checkpoint.",
+    )
+
+
+if __name__ == "__main__":
+    """Example usage:
+    python lama_inpaint.py \
+        --input_img FA_demo/FA1_dog.png \
+        --input_mask_glob "results/FA1_dog/mask*.png" \
+        --output_dir results \
+        --lama_config lama/configs/prediction/default.yaml \
+        --lama_ckpt big-lama 
+    """
+    parser = argparse.ArgumentParser()
+    setup_args(parser)
+    args = parser.parse_args(sys.argv[1:])
+    device = "cuda" if torch.cuda.is_available() else "cpu"
+
+    img_stem = Path(args.input_img).stem
+    mask_ps = sorted(glob.glob(args.input_mask_glob))
+    out_dir = Path(args.output_dir) / img_stem
+    out_dir.mkdir(parents=True, exist_ok=True)
+
+    img = load_img_to_array(args.input_img)
+    for mask_p in mask_ps:
+        mask = load_img_to_array(mask_p)
+        img_inpainted_p = out_dir / f"inpainted_with_{Path(mask_p).name}"
+        img_inpainted = inpaint_img_with_lama(
+            img, mask, args.lama_config, args.lama_ckpt, device=device)
+        save_array_to_img(img_inpainted, img_inpainted_p)

remove_anything.py

@@ -0,0 +1,132 @@
+import torch
+import sys
+import argparse
+import numpy as np
+from pathlib import Path
+from matplotlib import pyplot as plt
+
+from sam_segment import predict_masks_with_sam
+from lama_inpaint import inpaint_img_with_lama
+from utils import load_img_to_array, save_array_to_img, dilate_mask, \
+    show_mask, show_points, get_clicked_point
+
+
+def setup_args(parser):
+    parser.add_argument(
+        "--input_img", type=str, required=True,
+        help="Path to a single input img",
+    )
+    parser.add_argument(
+        "--coords_type", type=str, required=True,
+        default="key_in", choices=["click", "key_in"], 
+        help="The way to select coords",
+    )
+    parser.add_argument(
+        "--point_coords", type=float, nargs='+', required=True,
+        help="The coordinate of the point prompt, [coord_W coord_H].",
+    )
+    parser.add_argument(
+        "--point_labels", type=int, nargs='+', required=True,
+        help="The labels of the point prompt, 1 or 0.",
+    )
+    parser.add_argument(
+        "--dilate_kernel_size", type=int, default=None,
+        help="Dilate kernel size. Default: None",
+    )
+    parser.add_argument(
+        "--output_dir", type=str, required=True,
+        help="Output path to the directory with results.",
+    )
+    parser.add_argument(
+        "--sam_model_type", type=str,
+        default="vit_h", choices=['vit_h', 'vit_l', 'vit_b', 'vit_t'],
+        help="The type of sam model to load. Default: 'vit_h"
+    )
+    parser.add_argument(
+        "--sam_ckpt", type=str, required=True,
+        help="The path to the SAM checkpoint to use for mask generation.",
+    )
+    parser.add_argument(
+        "--lama_config", type=str,
+        default="./lama/configs/prediction/default.yaml",
+        help="The path to the config file of lama model. "
+             "Default: the config of big-lama",
+    )
+    parser.add_argument(
+        "--lama_ckpt", type=str, required=True,
+        help="The path to the lama checkpoint.",
+    )
+
+
+if __name__ == "__main__":
+    """Example usage:
+    python remove_anything.py \
+        --input_img FA_demo/FA1_dog.png \
+        --coords_type key_in \
+        --point_coords 750 500 \
+        --point_labels 1 \
+        --dilate_kernel_size 15 \
+        --output_dir ./results \
+        --sam_model_type "vit_h" \
+        --sam_ckpt sam_vit_h_4b8939.pth \
+        --lama_config lama/configs/prediction/default.yaml \
+        --lama_ckpt big-lama 
+    """
+    parser = argparse.ArgumentParser()
+    setup_args(parser)
+    args = parser.parse_args(sys.argv[1:])
+    device = "cuda" if torch.cuda.is_available() else "cpu"
+
+    if args.coords_type == "click":
+        latest_coords = get_clicked_point(args.input_img)
+    elif args.coords_type == "key_in":
+        latest_coords = args.point_coords
+    img = load_img_to_array(args.input_img)
+
+    masks, _, _ = predict_masks_with_sam(
+        img,
+        [latest_coords],
+        args.point_labels,
+        model_type=args.sam_model_type,
+        ckpt_p=args.sam_ckpt,
+        device=device,
+    )
+    masks = masks.astype(np.uint8) * 255
+
+    # dilate mask to avoid unmasked edge effect
+    if args.dilate_kernel_size is not None:
+        masks = [dilate_mask(mask, args.dilate_kernel_size) for mask in masks]
+
+    # visualize the segmentation results
+    img_stem = Path(args.input_img).stem
+    out_dir = Path(args.output_dir) / img_stem
+    out_dir.mkdir(parents=True, exist_ok=True)
+    for idx, mask in enumerate(masks):
+        # path to the results
+        mask_p = out_dir / f"mask_{idx}.png"
+        img_points_p = out_dir / f"with_points.png"
+        img_mask_p = out_dir / f"with_{Path(mask_p).name}"
+
+        # save the mask
+        save_array_to_img(mask, mask_p)
+
+        # save the pointed and masked image
+        dpi = plt.rcParams['figure.dpi']
+        height, width = img.shape[:2]
+        plt.figure(figsize=(width/dpi/0.77, height/dpi/0.77))
+        plt.imshow(img)
+        plt.axis('off')
+        show_points(plt.gca(), [latest_coords], args.point_labels,
+                    size=(width*0.04)**2)
+        plt.savefig(img_points_p, bbox_inches='tight', pad_inches=0)
+        show_mask(plt.gca(), mask, random_color=False)
+        plt.savefig(img_mask_p, bbox_inches='tight', pad_inches=0)
+        plt.close()
+
+    # inpaint the masked image
+    for idx, mask in enumerate(masks):
+        mask_p = out_dir / f"mask_{idx}.png"
+        img_inpainted_p = out_dir / f"inpainted_with_{Path(mask_p).name}"
+        img_inpainted = inpaint_img_with_lama(
+            img, mask, args.lama_config, args.lama_ckpt, device=device)
+        save_array_to_img(img_inpainted, img_inpainted_p)

replace_anything.py

@@ -0,0 +1,136 @@
+import cv2
+import sys
+import argparse
+import numpy as np
+import torch
+from pathlib import Path
+from matplotlib import pyplot as plt
+from typing import Any, Dict, List
+from sam_segment import predict_masks_with_sam
+from stable_diffusion_inpaint import replace_img_with_sd
+from utils import load_img_to_array, save_array_to_img, dilate_mask, \
+    show_mask, show_points, get_clicked_point
+
+
+def setup_args(parser):
+    parser.add_argument(
+        "--input_img", type=str, required=True,
+        help="Path to a single input img",
+    )
+    parser.add_argument(
+        "--coords_type", type=str, required=True,
+        default="key_in", choices=["click", "key_in"], 
+        help="The way to select coords",
+    )
+    parser.add_argument(
+        "--point_coords", type=float, nargs='+', required=True,
+        help="The coordinate of the point prompt, [coord_W coord_H].",
+    )
+    parser.add_argument(
+        "--point_labels", type=int, nargs='+', required=True,
+        help="The labels of the point prompt, 1 or 0.",
+    )
+    parser.add_argument(
+        "--text_prompt", type=str, required=True,
+        help="Text prompt",
+    )
+    parser.add_argument(
+        "--dilate_kernel_size", type=int, default=None,
+        help="Dilate kernel size. Default: None",
+    )
+    parser.add_argument(
+        "--output_dir", type=str, required=True,
+        help="Output path to the directory with results.",
+    )
+    parser.add_argument(
+        "--sam_model_type", type=str,
+        default="vit_h", choices=['vit_h', 'vit_l', 'vit_b', 'vit_t'],
+        help="The type of sam model to load. Default: 'vit_h"
+    )
+    parser.add_argument(
+        "--sam_ckpt", type=str, required=True,
+        help="The path to the SAM checkpoint to use for mask generation.",
+    )
+    parser.add_argument(
+        "--seed", type=int,
+        help="Specify seed for reproducibility.",
+    )
+    parser.add_argument(
+        "--deterministic", action="store_true",
+        help="Use deterministic algorithms for reproducibility.",
+    )
+
+
+
+if __name__ == "__main__":
+    """Example usage:
+    python replace_anything.py \
+        --input_img ./example/replace-anything/dog.png \
+        --coords_type key_in \
+        --point_coords 750 500 \
+        --point_labels 1 \
+        --text_prompt "sit on the swing" \
+        --output_dir ./results \
+        --sam_model_type "vit_h" \
+        --sam_ckpt ./pretrained_models/sam_vit_h_4b8939.pth
+    """
+    parser = argparse.ArgumentParser()
+    setup_args(parser)
+    args = parser.parse_args(sys.argv[1:])
+    device = "cuda" if torch.cuda.is_available() else "cpu"
+
+    if args.coords_type == "click":
+        latest_coords = get_clicked_point(args.input_img)
+    elif args.coords_type == "key_in":
+        latest_coords = args.point_coords
+    img = load_img_to_array(args.input_img)
+
+    masks, _, _ = predict_masks_with_sam(
+        img,
+        [latest_coords],
+        args.point_labels,
+        model_type=args.sam_model_type,
+        ckpt_p=args.sam_ckpt,
+        device=device,
+    )
+    masks = masks.astype(np.uint8) * 255
+
+    # dilate mask to avoid unmasked edge effect
+    if args.dilate_kernel_size is not None:
+        masks = [dilate_mask(mask, args.dilate_kernel_size) for mask in masks]
+
+    # visualize the segmentation results
+    img_stem = Path(args.input_img).stem
+    out_dir = Path(args.output_dir) / img_stem
+    out_dir.mkdir(parents=True, exist_ok=True)
+    for idx, mask in enumerate(masks):
+        # path to the results
+        mask_p = out_dir / f"mask_{idx}.png"
+        img_points_p = out_dir / f"with_points.png"
+        img_mask_p = out_dir / f"with_{Path(mask_p).name}"
+
+        # save the mask
+        save_array_to_img(mask, mask_p)
+
+        # save the pointed and masked image
+        dpi = plt.rcParams['figure.dpi']
+        height, width = img.shape[:2]
+        plt.figure(figsize=(width/dpi/0.77, height/dpi/0.77))
+        plt.imshow(img)
+        plt.axis('off')
+        show_points(plt.gca(), [latest_coords], args.point_labels,
+                    size=(width*0.04)**2)
+        plt.savefig(img_points_p, bbox_inches='tight', pad_inches=0)
+        show_mask(plt.gca(), mask, random_color=False)
+        plt.savefig(img_mask_p, bbox_inches='tight', pad_inches=0)
+        plt.close()
+
+    # fill the masked image
+    for idx, mask in enumerate(masks):
+        if args.seed is not None:
+            torch.manual_seed(args.seed)
+        mask_p = out_dir / f"mask_{idx}.png"
+        img_replaced_p = out_dir / f"replaced_with_{Path(mask_p).name}"
+        img_replaced = replace_img_with_sd(
+            img, mask, args.text_prompt, device=device)
+        save_array_to_img(img_replaced, img_replaced_p)

requirements.txt

@@ -0,0 +1,33 @@
+gradio
+
+torch
+torchvision
+torchaudio
+segment_anything
+diffusers
+transformers
+accelerate
+scipy
+safetensors
+
+# lama
+pyyaml
+tqdm
+numpy
+easydict==1.9.0
+scikit-image==0.17.2
+scikit-learn==0.24.2
+opencv-python
+tensorflow
+joblib
+matplotlib
+pandas
+albumentations==0.5.2
+hydra-core==1.1.0
+pytorch-lightning==1.2.9
+tabulate
+kornia==0.5.0
+webdataset
+packaging
+scikit-learn==0.24.2
+wldhx.yadisk-direct

sam_segment.py

@@ -0,0 +1,133 @@
+import sys
+import argparse
+import numpy as np
+from pathlib import Path
+from matplotlib import pyplot as plt
+from typing import Any, Dict, List
+import torch
+
+from segment_anything import SamPredictor, sam_model_registry
+from utils import load_img_to_array, save_array_to_img, dilate_mask, \
+    show_mask, show_points
+
+
+def predict_masks_with_sam(
+        img: np.ndarray,
+        point_coords: List[List[float]],
+        point_labels: List[int],
+        model_type: str,
+        ckpt_p: str,
+        device="cuda"
+):
+    point_coords = np.array(point_coords)
+    point_labels = np.array(point_labels)
+    sam = sam_model_registry[model_type](checkpoint=ckpt_p)
+    sam.to(device=device)
+    predictor = SamPredictor(sam)
+
+    predictor.set_image(img)
+    masks, scores, logits = predictor.predict(
+        point_coords=point_coords,
+        point_labels=point_labels,
+        multimask_output=True,
+    )
+    return masks, scores, logits
+
+
+def build_sam_model(model_type: str, ckpt_p: str, device="cuda"):
+    sam = sam_model_registry[model_type](checkpoint=ckpt_p)
+    sam.to(device=device)
+    predictor = SamPredictor(sam)
+    return predictor
+
+
+
+def setup_args(parser):
+    parser.add_argument(
+        "--input_img", type=str, required=True,
+        help="Path to a single input img",
+    )
+    parser.add_argument(
+        "--point_coords", type=float, nargs='+', required=True,
+        help="The coordinate of the point prompt, [coord_W coord_H].",
+    )
+    parser.add_argument(
+        "--point_labels", type=int, nargs='+', required=True,
+        help="The labels of the point prompt, 1 or 0.",
+    )
+    parser.add_argument(
+        "--dilate_kernel_size", type=int, default=None,
+        help="Dilate kernel size. Default: None",
+    )
+    parser.add_argument(
+        "--output_dir", type=str, required=True,
+        help="Output path to the directory with results.",
+    )
+    parser.add_argument(
+        "--sam_model_type", type=str,
+        default="vit_h", choices=['vit_h', 'vit_l', 'vit_b'],
+        help="The type of sam model to load. Default: 'vit_h"
+    )
+    parser.add_argument(
+        "--sam_ckpt", type=str, required=True,
+        help="The path to the SAM checkpoint to use for mask generation.",
+    )
+
+
+if __name__ == "__main__":
+    """Example usage:
+    python sam_segment.py \
+        --input_img FA_demo/FA1_dog.png \
+        --point_coords 750 500 \
+        --point_labels 1 \
+        --dilate_kernel_size 15 \
+        --output_dir ./results \
+        --sam_model_type "vit_h" \
+        --sam_ckpt sam_vit_h_4b8939.pth
+    """
+    parser = argparse.ArgumentParser()
+    setup_args(parser)
+    args = parser.parse_args(sys.argv[1:])
+    device = "cuda" if torch.cuda.is_available() else "cpu"
+
+    img = load_img_to_array(args.input_img)
+
+    masks, _, _ = predict_masks_with_sam(
+        img,
+        [args.point_coords],
+        args.point_labels,
+        model_type=args.sam_model_type,
+        ckpt_p=args.sam_ckpt,
+        device=device,
+    )
+    masks = masks.astype(np.uint8) * 255
+
+    # dilate mask to avoid unmasked edge effect
+    if args.dilate_kernel_size is not None:
+        masks = [dilate_mask(mask, args.dilate_kernel_size) for mask in masks]
+
+    # visualize the segmentation results
+    img_stem = Path(args.input_img).stem
+    out_dir = Path(args.output_dir) / img_stem
+    out_dir.mkdir(parents=True, exist_ok=True)
+    for idx, mask in enumerate(masks):
+        # path to the results
+        mask_p = out_dir / f"mask_{idx}.png"
+        img_points_p = out_dir / f"with_points.png"
+        img_mask_p = out_dir / f"with_{Path(mask_p).name}"
+
+        # save the mask
+        save_array_to_img(mask, mask_p)
+
+        # save the pointed and masked image
+        dpi = plt.rcParams['figure.dpi']
+        height, width = img.shape[:2]
+        plt.figure(figsize=(width/dpi/0.77, height/dpi/0.77))
+        plt.imshow(img)
+        plt.axis('off')
+        show_points(plt.gca(), [args.point_coords], args.point_labels,
+                    size=(width*0.04)**2)
+        plt.savefig(img_points_p, bbox_inches='tight', pad_inches=0)
+        show_mask(plt.gca(), mask, random_color=False)
+        plt.savefig(img_mask_p, bbox_inches='tight', pad_inches=0)
+        plt.close()

stable_diffusion_inpaint.py

@@ -0,0 +1,121 @@
+import os
+import sys
+import glob
+import argparse
+import torch
+import numpy as np
+import PIL.Image as Image
+from pathlib import Path
+from diffusers import StableDiffusionInpaintPipeline
+from utils.mask_processing import crop_for_filling_pre, crop_for_filling_post
+from utils.crop_for_replacing import recover_size, resize_and_pad
+from utils import load_img_to_array, save_array_to_img
+
+
+def fill_img_with_sd(
+    img: np.ndarray, mask: np.ndarray, text_prompt: str, device="cuda"
+):
+    pipe = StableDiffusionInpaintPipeline.from_pretrained(
+        "stabilityai/stable-diffusion-2-inpainting",
+        torch_dtype=torch.float32,
+    ).to(device)
+    img_crop, mask_crop = crop_for_filling_pre(img, mask)
+    img_crop_filled = pipe(
+        prompt=text_prompt,
+        image=Image.fromarray(img_crop),
+        mask_image=Image.fromarray(mask_crop),
+    ).images[0]
+    img_filled = crop_for_filling_post(img, mask, np.array(img_crop_filled))
+    return img_filled
+
+
+def replace_img_with_sd(
+    img: np.ndarray, mask: np.ndarray, text_prompt: str, step: int = 50, device="cuda"
+):
+    pipe = StableDiffusionInpaintPipeline.from_pretrained(
+        "stabilityai/stable-diffusion-2-inpainting",
+        torch_dtype=torch.float32,
+    ).to(device)
+    img_padded, mask_padded, padding_factors = resize_and_pad(img, mask)
+    img_padded = pipe(
+        prompt=text_prompt,
+        image=Image.fromarray(img_padded),
+        mask_image=Image.fromarray(255 - mask_padded),
+        num_inference_steps=step,
+    ).images[0]
+    height, width, _ = img.shape
+    img_resized, mask_resized = recover_size(
+        np.array(img_padded), mask_padded, (height, width), padding_factors
+    )
+    mask_resized = np.expand_dims(mask_resized, -1) / 255
+    img_resized = img_resized * (1 - mask_resized) + img * mask_resized
+    return img_resized
+
+
+def setup_args(parser):
+    parser.add_argument(
+        "--input_img",
+        type=str,
+        required=True,
+        help="Path to a single input img",
+    )
+    parser.add_argument(
+        "--text_prompt",
+        type=str,
+        required=True,
+        help="Text prompt",
+    )
+    parser.add_argument(
+        "--input_mask_glob",
+        type=str,
+        required=True,
+        help="Glob to input masks",
+    )
+    parser.add_argument(
+        "--output_dir",
+        type=str,
+        required=True,
+        help="Output path to the directory with results.",
+    )
+    parser.add_argument(
+        "--seed",
+        type=int,
+        help="Specify seed for reproducibility.",
+    )
+    parser.add_argument(
+        "--deterministic",
+        action="store_true",
+        help="Use deterministic algorithms for reproducibility.",
+    )
+
+
+if __name__ == "__main__":
+    """Example usage:
+    python lama_inpaint.py \
+        --input_img FA_demo/FA1_dog.png \
+        --input_mask_glob "results/FA1_dog/mask*.png" \
+        --text_prompt "a teddy bear on a bench" \
+        --output_dir results 
+    """
+    parser = argparse.ArgumentParser()
+    setup_args(parser)
+    args = parser.parse_args(sys.argv[1:])
+    device = "cuda" if torch.cuda.is_available() else "cpu"
+
+    if args.deterministic:
+        os.environ["CUBLAS_WORKSPACE_CONFIG"] = ":4096:8"
+        torch.use_deterministic_algorithms(True)
+
+    img_stem = Path(args.input_img).stem
+    mask_ps = sorted(glob.glob(args.input_mask_glob))
+    out_dir = Path(args.output_dir) / img_stem
+    out_dir.mkdir(parents=True, exist_ok=True)
+
+    img = load_img_to_array(args.input_img)
+    for mask_p in mask_ps:
+        if args.seed is not None:
+            torch.manual_seed(args.seed)
+        mask = load_img_to_array(mask_p)
+        img_filled_p = out_dir / f"filled_with_{Path(mask_p).name}"
+        img_filled = fill_img_with_sd(img, mask, args.text_prompt, device=device)
+        save_array_to_img(img_filled, img_filled_p)