Refine ImageBind-SAM playground (IDEA-Research#273)

* update imagebind sam

* refine readme

README.md

@@ -14,6 +14,7 @@ We are very willing to **help everyone share and promote new projects** based on
 The **core idea** behind this project is to **combine the strengths of different models in order to build a very powerful pipeline for solving complex problems**. And it's worth mentioning that this is a workflow for combining strong expert models, where **all parts can be used separately or in combination, and can be replaced with any similar but different models (like replacing Grounding DINO with GLIP or other detectors / replacing Stable-Diffusion with ControlNet or GLIGEN/ Combining with ChatGPT)**.
 
 **🍇 Updates**
+- **`2023/05/23`**: Support `Image-Referring-Segment`, `Audio-Referring-Segment` and `Text-Referring-Segment` in [ImageBind-SAM](./playground/ImageBind_SAM/).
 - **`2023/05/16`**: Release [ImageBind-SAM](./playground/ImageBind_SAM/) simple demo which aims to segment with different modalities.
 - **`2023/05/15`**: Release [LaMa](./playground/LaMa/) and [RePaint](./playground/RePaint/) demo, thanks for nice tips by [Tao Yu](https://github.com/geekyutao).
 - **`2023/05/14`**: Release [PaintByExample](./playground/PaintByExample/) demo with SAM.

playground/ImageBind_SAM/README.md

@@ -12,6 +12,10 @@ This basic idea is followed with [IEA: Image Editing Anything](https://github.co
 ## Table of contents
 - [Installation](#installation)
 - [ImageBind-SAM Demo](#run-the-demo)
+- [Audio Referring Segment](#run-audio-referring-segment-demo)
+- [Text Referring Segment](#run-text-referring-segment-demo)
+- [Image Referring Segment](#run-image-referring-segmentation-demo)
+
 
 
 ## Installation
@@ -45,7 +49,29 @@ We implement `Text Seg` and `Audio Seg` in this demo, the generate masks will be
 |:----:|:----:|:----:|
 | ![](./.assets/car_image.jpg) | [car audio](./.assets/car_audio.wav) | ![](https://github.com/IDEA-Research/detrex-storage/blob/main/assets/grounded_sam/imagebind_sam/audio_sam_merged_mask_new.jpg?raw=true) |
 | ![](./.assets/car_image.jpg) | "A car" | ![](https://github.com/IDEA-Research/detrex-storage/blob/main/assets/grounded_sam/imagebind_sam/text_sam_merged_mask.jpg?raw=true) |
+| ![](./.assets/car_image.jpg) | <div style="text-align: center"> <img src="https://github.com/IDEA-Research/detrex-storage/blob/main/assets/grounded_sam/imagebind_sam/referring_car_image.jpg?raw=true" width=55%></div> | ![](https://github.com/IDEA-Research/detrex-storage/blob/main/assets/grounded_sam/imagebind_sam/image_referring_sam_merged_mask.jpg?raw=true) |
+
 
 </div>
 
-By setting different threshold may influence a lot on the final results.
+By setting different threshold may influence a lot on the final results.
+
+## Run image referring segmentation demo
+```bash
+# download the referring image
+cd .assets
+wget https://github.com/IDEA-Research/detrex-storage/releases/download/grounded-sam-storage/referring_car_image.jpg
+cd ..
+
+python image_referring_seg_demo.py
+```
+
+## Run audio referring segmentation demo
+```bash
+python audio_referring_seg_demo.py
+```
+
+## Run text referring segmentation demo
+```bash
+python text_referring_seg_demo.py
+```

playground/ImageBind_SAM/audio_referring_seg_demo.py

@@ -0,0 +1,97 @@
+import data
+import cv2
+import torch
+from PIL import Image, ImageDraw
+from tqdm import tqdm
+from models import imagebind_model
+from models.imagebind_model import ModalityType
+
+from segment_anything import build_sam, SamAutomaticMaskGenerator
+
+from utils import (
+    segment_image, 
+    convert_box_xywh_to_xyxy,
+    get_indices_of_values_above_threshold,
+)
+
+
+device = "cuda" if torch.cuda.is_available() else "cpu"
+
+
+"""
+Step 1: Instantiate model
+"""
+# Segment Anything
+mask_generator = SamAutomaticMaskGenerator(
+    build_sam(checkpoint=".checkpoints/sam_vit_h_4b8939.pth").to(device),
+    points_per_side=16,
+)
+
+# ImageBind
+bind_model = imagebind_model.imagebind_huge(pretrained=True)
+bind_model.eval()
+bind_model.to(device)
+
+
+"""
+Step 2: Generate auto masks with SAM
+"""
+image_path = ".assets/car_image.jpg"
+image = cv2.imread(image_path)
+image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
+masks = mask_generator.generate(image)
+
+
+"""
+Step 3: Get cropped images based on mask and box
+"""
+cropped_boxes = []
+image = Image.open(image_path)
+for mask in tqdm(masks):
+    cropped_boxes.append(segment_image(image, mask["segmentation"]).crop(convert_box_xywh_to_xyxy(mask["bbox"])))
+
+
+"""
+Step 4: Run ImageBind model to get similarity between cropped image and different modalities
+"""
+def retriev_vision_and_audio(elements, audio_list):
+    inputs = {
+        ModalityType.VISION: data.load_and_transform_vision_data_from_pil_image(elements, device),
+        ModalityType.AUDIO: data.load_and_transform_audio_data(audio_list, device),
+    }
+    with torch.no_grad():
+        embeddings = bind_model(inputs)
+    vision_audio = torch.softmax(embeddings[ModalityType.VISION] @ embeddings[ModalityType.AUDIO].T, dim=0),
+    return vision_audio 
+
+vision_audio_result = retriev_vision_and_audio(cropped_boxes, [".assets/car_audio.wav"])
+
+
+"""
+Step 5: Merge the top similarity masks to get the final mask and save the merged mask
+
+This is the audio retrival result
+"""
+
+# get highest similar mask with threshold
+# result[0] shape: [113, 1]
+threshold = 0.025
+index = get_indices_of_values_above_threshold(vision_audio_result[0], threshold)
+
+segmentation_masks = []
+for seg_idx in index:
+    segmentation_mask_image = Image.fromarray(masks[seg_idx]["segmentation"].astype('uint8') * 255)
+    segmentation_masks.append(segmentation_mask_image)
+
+original_image = Image.open(image_path)
+overlay_image = Image.new('RGBA', image.size, (0, 0, 0, 255))
+overlay_color = (255, 255, 255, 0)
+
+draw = ImageDraw.Draw(overlay_image)
+for segmentation_mask_image in segmentation_masks:
+    draw.bitmap((0, 0), segmentation_mask_image, fill=overlay_color)
+
+# return Image.alpha_composite(original_image.convert('RGBA'), overlay_image) 
+mask_image = overlay_image.convert("RGB")
+mask_image.save("./audio_sam_merged_mask.jpg")
+

playground/ImageBind_SAM/image_referring_seg_demo.py

@@ -0,0 +1,110 @@
+import data
+import cv2
+import torch
+from PIL import Image, ImageDraw
+from tqdm import tqdm
+from models import imagebind_model
+from models.imagebind_model import ModalityType
+
+from segment_anything import build_sam, SamAutomaticMaskGenerator
+
+from utils import (
+    segment_image, 
+    convert_box_xywh_to_xyxy,
+    get_indices_of_values_above_threshold,
+)
+
+
+device = "cuda" if torch.cuda.is_available() else "cpu"
+
+
+"""
+Step 1: Instantiate model
+"""
+# Segment Anything
+mask_generator = SamAutomaticMaskGenerator(
+    build_sam(checkpoint=".checkpoints/sam_vit_h_4b8939.pth").to(device),
+    points_per_side=16,
+)
+
+# ImageBind
+bind_model = imagebind_model.imagebind_huge(pretrained=True)
+bind_model.eval()
+bind_model.to(device)
+
+
+"""
+Step 2: Generate auto masks with SAM
+"""
+image_path = ".assets/car_image.jpg"
+image = cv2.imread(image_path)
+image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
+masks = mask_generator.generate(image)
+
+
+"""
+Step 3: Get cropped images based on mask and box
+"""
+cropped_boxes = []
+image = Image.open(image_path)
+for mask in tqdm(masks):
+    cropped_boxes.append(segment_image(image, mask["segmentation"]).crop(convert_box_xywh_to_xyxy(mask["bbox"])))
+
+
+"""
+Step 4: Run ImageBind model to get similarity between cropped image and different modalities
+"""
+# load referring image
+referring_image_path = ".assets/referring_car_image.jpg"
+referring_image = Image.open(referring_image_path)
+
+image_list = []
+image_list += cropped_boxes
+image_list.append(referring_image)
+
+def retriev_vision_and_vision(elements):
+    inputs = {
+        ModalityType.VISION: data.load_and_transform_vision_data_from_pil_image(elements, device),
+    }
+    with torch.no_grad():
+        embeddings = bind_model(inputs)
+
+    # cropped box region embeddings
+    cropped_box_embeddings = embeddings[ModalityType.VISION][:-1, :]
+    referring_image_embeddings = embeddings[ModalityType.VISION][-1, :]
+
+    vision_referring_result = torch.softmax(cropped_box_embeddings @ referring_image_embeddings.T, dim=0),
+    return vision_referring_result  # [113, 1]
+
+
+vision_referring_result = retriev_vision_and_vision(image_list)
+
+
+"""
+Step 5: Merge the top similarity masks to get the final mask and save the merged mask
+
+Image / Text mask
+"""
+
+# get highest similar mask with threshold
+# result[0] shape: [113, 1]
+threshold = 0.017
+index = get_indices_of_values_above_threshold(vision_referring_result[0], threshold)
+
+
+segmentation_masks = []
+for seg_idx in index:
+    segmentation_mask_image = Image.fromarray(masks[seg_idx]["segmentation"].astype('uint8') * 255)
+    segmentation_masks.append(segmentation_mask_image)
+
+original_image = Image.open(image_path)
+overlay_image = Image.new('RGBA', image.size, (0, 0, 0, 255))
+overlay_color = (255, 255, 255, 0)
+
+draw = ImageDraw.Draw(overlay_image)
+for segmentation_mask_image in segmentation_masks:
+    draw.bitmap((0, 0), segmentation_mask_image, fill=overlay_color)
+
+# return Image.alpha_composite(original_image.convert('RGBA'), overlay_image) 
+mask_image = overlay_image.convert("RGB")
+mask_image.save("./image_referring_sam_merged_mask.jpg")

playground/ImageBind_SAM/text_referring_seg_demo.py

@@ -0,0 +1,97 @@
+import data
+import cv2
+import torch
+from PIL import Image, ImageDraw
+from tqdm import tqdm
+from models import imagebind_model
+from models.imagebind_model import ModalityType
+
+from segment_anything import build_sam, SamAutomaticMaskGenerator
+
+from utils import (
+    segment_image, 
+    convert_box_xywh_to_xyxy,
+    get_indices_of_values_above_threshold,
+)
+
+
+device = "cuda" if torch.cuda.is_available() else "cpu"
+
+
+"""
+Step 1: Instantiate model
+"""
+# Segment Anything
+mask_generator = SamAutomaticMaskGenerator(
+    build_sam(checkpoint=".checkpoints/sam_vit_h_4b8939.pth").to(device),
+    points_per_side=16,
+)
+
+# ImageBind
+bind_model = imagebind_model.imagebind_huge(pretrained=True)
+bind_model.eval()
+bind_model.to(device)
+
+
+"""
+Step 2: Generate auto masks with SAM
+"""
+image_path = ".assets/car_image.jpg"
+image = cv2.imread(image_path)
+image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
+masks = mask_generator.generate(image)
+
+
+"""
+Step 3: Get cropped images based on mask and box
+"""
+cropped_boxes = []
+image = Image.open(image_path)
+for mask in tqdm(masks):
+    cropped_boxes.append(segment_image(image, mask["segmentation"]).crop(convert_box_xywh_to_xyxy(mask["bbox"])))
+
+
+"""
+Step 4: Run ImageBind model to get similarity between cropped image and different modalities
+"""
+def retriev_vision_and_text(elements, text_list):
+    inputs = {
+        ModalityType.VISION: data.load_and_transform_vision_data_from_pil_image(elements, device),
+        ModalityType.TEXT: data.load_and_transform_text(text_list, device),
+    }
+    with torch.no_grad():
+        embeddings = bind_model(inputs)
+    vision_audio = torch.softmax(embeddings[ModalityType.VISION] @ embeddings[ModalityType.TEXT].T, dim=0),
+    return vision_audio  # [113, 1]
+
+
+vision_text_result = retriev_vision_and_text(cropped_boxes, ["A car"] )
+
+
+"""
+Step 5: Merge the top similarity masks to get the final mask and save the merged mask
+
+Image / Text mask
+"""
+
+# get highest similar mask with threshold
+# result[0] shape: [113, 1]
+threshold = 0.05
+index = get_indices_of_values_above_threshold(vision_text_result[0], threshold)
+
+segmentation_masks = []
+for seg_idx in index:
+    segmentation_mask_image = Image.fromarray(masks[seg_idx]["segmentation"].astype('uint8') * 255)
+    segmentation_masks.append(segmentation_mask_image)
+
+original_image = Image.open(image_path)
+overlay_image = Image.new('RGBA', image.size, (0, 0, 0, 255))
+overlay_color = (255, 255, 255, 0)
+
+draw = ImageDraw.Draw(overlay_image)
+for segmentation_mask_image in segmentation_masks:
+    draw.bitmap((0, 0), segmentation_mask_image, fill=overlay_color)
+
+# return Image.alpha_composite(original_image.convert('RGBA'), overlay_image) 
+mask_image = overlay_image.convert("RGB")
+mask_image.save("./text_sam_merged_mask.jpg")