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

scripts/debug/dump_logs.py

@@ -0,0 +1,68 @@
+import cv2
+import pickle
+from pathlib import Path
+
+# TemporalTracker already saved the video chunks? No.
+# I will use fast_verify.py's frames but run process_pages directly and print all its output.
+import fast_verify
+from youtube_tab_to_pdf import extract_unique_scroll
+
+# Actually, I will just write a wrapper around ScoreExtractor to print to file
+import sys
+
+def main():
+    cap = cv2.VideoCapture("output/サカナクション／新宝島(エレキギターTAB) 難易度★★★ sakanaction shintakarajima.mp4")
+    # Quick dynamic crop
+    ret, initial = cap.read()
+    scale = 1280 / initial.shape[1]
+    
+    from youtube_tab_to_pdf import _find_white_tab_strip
+    crop_top, crop_bottom = 0, int(initial.shape[0] * scale)
+    
+    cap.set(cv2.CAP_PROP_POS_FRAMES, 500)
+    ret, check_frame = cap.read()
+    if ret:
+        resized_check = cv2.resize(check_frame, (1280, int(check_frame.shape[0] * scale)))
+        bounds = _find_white_tab_strip(resized_check)
+        if bounds:
+            crop_top, crop_bottom = bounds
+            
+    cap.set(cv2.CAP_PROP_POS_FRAMES, 0)
+    # We don't want to load ALL 15000 frames into memory. Use TemporalTracker directly!
+    from video_cv_tracker import TemporalTracker
+    tracker = TemporalTracker(diff_threshold=0.05)
+    
+    count = 0
+    cap.set(cv2.CAP_PROP_POS_FRAMES, 0)
+    
+    while True:
+        ret, frame = cap.read()
+        if not ret: break
+        if count % 4 == 0:
+            resized = cv2.resize(frame, (1280, int(frame.shape[0] * scale)))
+            tracker.process_frame(resized[crop_top:crop_bottom, :])
+        count += 1
+        
+    unique_pages = tracker.get_unique_pages()
+    print(f"Got {len(unique_pages)} unique pages from tracker.")
+    
+    from score_extractor import ScoreExtractor
+    extractor = ScoreExtractor()
+    
+    # We will hook print
+    original_print = print
+    with open("score_log.txt", "w") as f:
+        def my_print(*args, **kwargs):
+            text = " ".join(map(str, args))
+            f.write(text + "\n")
+            original_print(*args, **kwargs)
+        
+        import builtins
+        builtins.print = my_print
+        
+        extractor.process_pages(unique_pages)
+        
+        builtins.print = original_print
+
+if __name__ == "__main__":
+    main()