DiffusionDet icon indicating copy to clipboard operation
DiffusionDet copied to clipboard

Too many boxes when using demo.py with --video-input video.mp4

Open BrainPotter opened this issue 1 year ago • 2 comments

Hello! Thanks for your great working! I run demo.py with --video-input. But there are too many boxes in the generated results. Please tell me what I should do. 微信图片_20240509170518 I Adjust the threshold value in the detector.py. This problem cannot be solved. image I also run demo.py with --input image.jpg, the results seems correct. 000095_L_threshold_0 8 I really want to know why? PS: trained on COCO and LVIS

BrainPotter avatar May 09 '24 09:05 BrainPotter

@BrainPotter Hello, i have same issue before. I checked the predictor.py and i found that on video there is no score_threshold filtering. you add it like this.

Image

MusyafaRT avatar Jan 21 '25 08:01 MusyafaRT

Thank you very much!

BrainPotter avatar Feb 28 '25 06:02 BrainPotter