Candle
commited on
Commit
·
3927f54
1
Parent(s):
1b7edb8
stuff
Browse files- detect_scene.py +14 -2
detect_scene.py
CHANGED
|
@@ -8,6 +8,7 @@ import re
|
|
| 8 |
# SCENE_CUT_THRESHOLD = 0.09
|
| 9 |
K = 3 # Number of cuts to detect
|
| 10 |
MIN_DURATION_FRAMES = 2
|
|
|
|
| 11 |
|
| 12 |
data_dir = Path("data/animations")
|
| 13 |
files = sorted(data_dir.glob("sample-*.webp"))
|
|
@@ -209,21 +210,32 @@ if __name__ == "__main__":
|
|
| 209 |
# Sort indices by prediction score (descending)
|
| 210 |
sorted_indices = valid_indices[np.argsort(valid_preds)[::-1]]
|
| 211 |
scene_change_indices = []
|
|
|
|
| 212 |
for idx in sorted_indices:
|
| 213 |
if all(abs(idx - prev) >= MIN_DURATION_FRAMES for prev in scene_change_indices):
|
| 214 |
scene_change_indices.append(int(idx))
|
|
|
|
| 215 |
if len(scene_change_indices) >= (K - 1):
|
| 216 |
break
|
| 217 |
|
| 218 |
-
|
|
|
|
|
|
|
|
|
|
| 219 |
# Save results to JSON (include threshold and predictions)
|
| 220 |
json_filename = file.parent / f"sample-{sample_num}.json"
|
| 221 |
with open(json_filename, "w") as f:
|
| 222 |
json.dump({
|
| 223 |
"num_frames": len(original_frames),
|
| 224 |
"scene_change_indices": scene_change_indices,
|
|
|
|
| 225 |
# "threshold": SCENE_CUT_THRESHOLD
|
| 226 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 227 |
}, f, indent=2)
|
| 228 |
# Save timeline JPG
|
| 229 |
timeline_filename = file.parent / f"sample-{sample_num}.timeline.jpg"
|
|
|
|
| 8 |
# SCENE_CUT_THRESHOLD = 0.09
|
| 9 |
K = 3 # Number of cuts to detect
|
| 10 |
MIN_DURATION_FRAMES = 2
|
| 11 |
+
MIN_CONFIDENCE = 0.02
|
| 12 |
|
| 13 |
data_dir = Path("data/animations")
|
| 14 |
files = sorted(data_dir.glob("sample-*.webp"))
|
|
|
|
| 210 |
# Sort indices by prediction score (descending)
|
| 211 |
sorted_indices = valid_indices[np.argsort(valid_preds)[::-1]]
|
| 212 |
scene_change_indices = []
|
| 213 |
+
scene_cut_confidences = []
|
| 214 |
for idx in sorted_indices:
|
| 215 |
if all(abs(idx - prev) >= MIN_DURATION_FRAMES for prev in scene_change_indices):
|
| 216 |
scene_change_indices.append(int(idx))
|
| 217 |
+
scene_cut_confidences.append(float(single_frame_pred[idx]))
|
| 218 |
if len(scene_change_indices) >= (K - 1):
|
| 219 |
break
|
| 220 |
|
| 221 |
+
# Check if any confidence is below MIN_CONFIDENCE
|
| 222 |
+
failed = any(conf < MIN_CONFIDENCE for conf in scene_cut_confidences)
|
| 223 |
+
|
| 224 |
+
print(f"File: {file.name}, Frames: {len(original_frames)}, Scene Changes: {len(scene_change_indices)}, Success: {not failed}")
|
| 225 |
# Save results to JSON (include threshold and predictions)
|
| 226 |
json_filename = file.parent / f"sample-{sample_num}.json"
|
| 227 |
with open(json_filename, "w") as f:
|
| 228 |
json.dump({
|
| 229 |
"num_frames": len(original_frames),
|
| 230 |
"scene_change_indices": scene_change_indices,
|
| 231 |
+
"scene_cut_confidences": scene_cut_confidences,
|
| 232 |
# "threshold": SCENE_CUT_THRESHOLD
|
| 233 |
+
"params": {
|
| 234 |
+
"k": K,
|
| 235 |
+
"min_duration_frames": MIN_DURATION_FRAMES,
|
| 236 |
+
"min_confidence": MIN_CONFIDENCE,
|
| 237 |
+
},
|
| 238 |
+
"success": not failed
|
| 239 |
}, f, indent=2)
|
| 240 |
# Save timeline JPG
|
| 241 |
timeline_filename = file.parent / f"sample-{sample_num}.timeline.jpg"
|