Spaces:
Runtime error
Runtime error
Commit History
-
Changed from yolov8n-world.pt to yolov8s-world.pt, add error handling
Verified signature9c5fb04 -
Update requirements.txt
Verified signature2130cf9 -
Limit frames, optimize code
Verified signatureSAM Resolution Problem: The original code was calling sam_model.predict(source=video_path, ...) which processes the entire video at SAM's default resolution (1024). I changed this to process individual frames with explicit imgsz=320 parameter. Inefficient Frame Processing: Your code was opening a new VideoCapture for each frame in the loop (cv2.VideoCapture(video_path).read()[1]), which is extremely inefficient. Missing Resolution Control for YOLO: Added imgsz=320 to the YOLO prediction to ensure consistent resolution. Box Scaling Issues: Removed unnecessary box scaling since we're working consistently in the target resolution. Memory Leaks: Fixed VideoCapture resource management7a3f21a -
Use mobile sam to reduce processing time
Verified signaturee5ea4ad -
remove supervision
Verified signature5ada4a5 -
Remove device argument, simplify model loading
Verified signature2974d2b -
Try to process individual frames to fix error
Verified signature801bb4d -
Use ultralytics import
Verified signature93c5c43 -
Create requirements.txt
Verified signatured31f23a -
Create app.py
Verified signature483fb8b
-
initial commit
Verified signature87560e7