feat: update Python processors and add utility scripts

- Update ASR, face, OCR, pose processors - Add release pre-flight check script - Add synonym generation, chunk processing scripts - Add face recognition, stamp search utilities
2026-04-30 15:07:49 +08:00
parent f4697396e4
commit 8f05a7c188
256 changed files with 60505 additions and 299 deletions
--- a/scripts/magnifying_glass_owl.py
+++ b/scripts/magnifying_glass_owl.py
@@ -0,0 +1,161 @@
+#!/opt/homebrew/bin/python3.11
+"""
+Magnifying Glass: OWL-ViT fine-grained stamp search
+Scans key frames with multiple stamp-related search terms.
+"""
+
+import os
+import cv2
+import json
+import glob
+from PIL import Image
+import torch
+from transformers import OwlViTProcessor, OwlViTForObjectDetection
+
+UUID = "384b0ff44aaaa1f1"
+BASE_DIR = f"output/{UUID}/magnifying_glass"
+RESULTS_DIR = f"output/{UUID}/magnifying_glass_owl"
+os.makedirs(RESULTS_DIR, exist_ok=True)
+
+print("🔬 Loading OWL-ViT model...")
+processor = OwlViTProcessor.from_pretrained("google/owlvit-base-patch32")
+model = OwlViTForObjectDetection.from_pretrained("google/owlvit-base-patch32")
+model.eval()
+
+# Comprehensive search terms for stamp detection
+SEARCH_TERMS = [
+    "postage stamp",
+    "stamp on envelope",
+    "stamp on paper",
+    "holding a stamp",
+    "envelope with stamp",
+    "letter with stamp",
+    "stamp collection",
+    "stamp album",
+    "rare stamp",
+    "British stamp",
+    "old stamp",
+    "small rectangular stamp",
+    "red stamp",
+    "blue stamp",
+    "stamp on document",
+    "envelope",
+    "letter",
+    "piece of paper",
+    "document",
+    "hand holding paper",
+]
+
+
+def detect_stamps(image_path, search_terms):
+    """Run OWL-ViT detection with multiple search terms"""
+    image = Image.open(image_path).convert("RGB")
+
+    all_detections = []
+
+    for term in search_terms:
+        inputs = processor(text=[[term]], images=image, return_tensors="pt")
+
+        with torch.no_grad():
+            outputs = model(**inputs)
+
+        # Use lower threshold for small objects
+        threshold = 0.05
+        target_sizes = torch.Tensor([image.size[::-1]])
+        results = processor.post_process_object_detection(
+            outputs=outputs, target_sizes=target_sizes, threshold=threshold
+        )
+
+        for score, label, box in zip(
+            results[0]["scores"], results[0]["labels"], results[0]["boxes"]
+        ):
+            if score > threshold:
+                all_detections.append(
+                    {
+                        "term": term,
+                        "score": float(score),
+                        "bbox": box.tolist(),
+                        "label": f"{term} ({score:.2f})",
+                    }
+                )
+
+    return all_detections
+
+
+def analyze_scene(scene_dir, scene_name):
+    """Analyze all frames in a scene"""
+    frames = sorted(glob.glob(os.path.join(scene_dir, "frame_*.jpg")))
+    print(f"\n🔍 Analyzing {scene_name}: {len(frames)} frames")
+
+    scene_results = []
+
+    for frame_path in frames:
+        frame_name = os.path.basename(frame_path)
+        sec = frame_name.replace("frame_", "").replace("s.jpg", "")
+
+        print(f"  Processing {sec}s...")
+        detections = detect_stamps(frame_path, SEARCH_TERMS)
+
+        if detections:
+            # Sort by score
+            detections.sort(key=lambda x: x["score"], reverse=True)
+            top_dets = detections[:10]  # Keep top 10
+
+            print(
+                f"    📍 Found {len(detections)} detections, top: {top_dets[0]['term']} ({top_dets[0]['score']:.2f})"
+            )
+
+            # Save annotated image
+            try:
+                img = cv2.imread(frame_path)
+                for det in top_dets:
+                    x1, y1, x2, y2 = map(int, det["bbox"])
+                    cv2.rectangle(img, (x1, y1), (x2, y2), (0, 255, 0), 2)
+                    cv2.putText(
+                        img,
+                        det["label"],
+                        (x1, y1 - 10),
+                        cv2.FONT_HERSHEY_SIMPLEX,
+                        0.5,
+                        (0, 255, 0),
+                        1,
+                    )
+
+                    # Save crop
+                    crop = img[y1:y2, x1:x2]
+                    if crop.size > 0:
+                        crop_name = (
+                            f"{scene_name}_{sec}s_{det['term'].replace(' ', '_')}.jpg"
+                        )
+                        cv2.imwrite(os.path.join(RESULTS_DIR, crop_name), crop)
+
+                ann_path = os.path.join(
+                    RESULTS_DIR, f"annotated_{scene_name}_{sec}s.jpg"
+                )
+                cv2.imwrite(ann_path, img)
+            except Exception as e:
+                print(f"    ⚠️  Save error: {e}")
+
+            scene_results.append({"frame": frame_name, "detections": top_dets})
+
+    return scene_results
+
+
+# Analyze all scenes
+all_results = {}
+scene_dirs = sorted(glob.glob(os.path.join(BASE_DIR, "*/")))
+print(f"📂 Found {len(scene_dirs)} scene directories")
+
+for scene_dir in scene_dirs:
+    scene_name = os.path.basename(os.path.dirname(scene_dir))
+    results = analyze_scene(scene_dir, scene_name)
+    if results:
+        all_results[scene_name] = results
+
+# Save results
+results_path = os.path.join(RESULTS_DIR, "detection_results.json")
+with open(results_path, "w") as f:
+    json.dump(all_results, f, indent=2)
+
+print(f"\n🏁 Done. Results saved to {results_path}")
+print(f"📁 Check {RESULTS_DIR} for annotated images and crops.")