feat: depth_masks.py script using Depth Anything V2

tools/depth_masks.py

@@ -0,0 +1,75 @@
+"""Depth Anything V2 mask generation script.
+
+Usage:
+    python tools/depth_masks.py --input video.mp4 --output masks_dir/
+
+Outputs one binary PNG per frame: frame_0000.png, frame_0001.png, …
+Foreground = white (255), background = black (0), via Otsu threshold on depth map.
+Requires: torch, transformers, opencv-python, Pillow
+"""
+import argparse
+import os
+import sys
+
+import cv2
+import numpy as np
+from PIL import Image
+from transformers import pipeline
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--input", required=True)
+    parser.add_argument("--output", required=True)
+    args = parser.parse_args()
+
+    os.makedirs(args.output, exist_ok=True)
+
+    import torch
+    device = "cuda" if torch.cuda.is_available() else "cpu"
+    print(f"Using device: {device}", flush=True)
+
+    pipe = pipeline(
+        "depth-estimation",
+        model="depth-anything/Depth-Anything-V2-Large-hf",
+        device=device,
+    )
+
+    cap = cv2.VideoCapture(args.input)
+    if not cap.isOpened():
+        print(f"ERROR: cannot open {args.input}", file=sys.stderr)
+        sys.exit(1)
+
+    total = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
+    idx = 0
+    while True:
+        ret, frame = cap.read()
+        if not ret:
+            break
+
+        pil_img = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
+        result = pipe(pil_img)
+        depth = np.array(result["depth"])  # float32 array
+
+        # Normalise to 0–255
+        d_min, d_max = depth.min(), depth.max()
+        if d_max > d_min:
+            depth_u8 = ((depth - d_min) / (d_max - d_min) * 255).astype(np.uint8)
+        else:
+            depth_u8 = np.zeros_like(depth, dtype=np.uint8)
+
+        # Otsu threshold: closer objects (higher depth value) = foreground
+        _, mask = cv2.threshold(depth_u8, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)
+
+        out_path = os.path.join(args.output, f"frame_{idx:04d}.png")
+        cv2.imwrite(out_path, mask)
+
+        idx += 1
+        print(f"frame {idx}/{total}", flush=True)
+
+    cap.release()
+    print("done", flush=True)
+
+
+if __name__ == "__main__":
+    main()