chore(docs): document ScoreExtractor tiling and refactor debug scripts (#563)

scripts/debug/debug_crash.py

@@ -0,0 +1,48 @@
+import cv2
+import pickle
+import traceback
+
+try:
+    with open('unique_pages.pkl', 'rb') as f:
+        unique_pages = pickle.load(f)
+except Exception:
+    import video_cv_tracker as tracker_lib
+    from youtube_tab_to_pdf import _find_white_tab_strip
+    tracker = tracker_lib.TemporalTracker(diff_threshold=0.05)
+    video = cv2.VideoCapture("output/shintakarajima.mp4")
+    
+    # Just read 100 frames
+    frames = []
+    fps_orig = video.get(cv2.CAP_PROP_FPS)
+    stride = max(1, int(fps_orig / 4.0))
+    count = 0
+    while len(frames) < 150:
+        ret, f = video.read()
+        if not ret: break
+        if count % stride == 0: frames.append(f)
+        count += 1
+    video.release()
+    
+    top, bottom = 0, frames[0].shape[0]
+    for f in frames[::10]:
+        b = _find_white_tab_strip(f)
+        if b:
+            top, bottom = b
+            break
+            
+    for f in frames:
+        tracker.process_frame(f[top:bottom, :])
+        
+    unique_pages = tracker.get_unique_pages()
+    with open('unique_pages.pkl', 'wb') as f:
+        pickle.dump(unique_pages, f)
+
+from score_extractor import ScoreExtractor
+ex = ScoreExtractor()
+try:
+    print(f"Running ScoreExtractor on {len(unique_pages)} pages...")
+    ex.process_pages(unique_pages)
+    print("Success!")
+except Exception as e:
+    print("CRASHED:")
+    traceback.print_exc()