search

ifzhang / ByteTrack

[ECCV 2022] ByteTrack: Multi-Object Tracking by Associating Every Detection Box
MIT License
4.8k stars 906 forks source link

TensorRT C++ inference results do not have bbox with custom dataset #255

Open zzh805780186 opened 2 years ago

zzh805780186 commented 2 years ago

I can use author's 'model_trt.engine' deploy 'palace.mp4' normally.

problem1

However, I generate 'bytetrack_s.onnx' , with my dataset,and make 'model_trt.engine', there have no bbox when inference. Who can tell me why? Help!

problem2

zzh805780186 commented 2 years ago

I find object detection box_prob is a very small value , and much less then prob_threshold. It means yolox can not detect any bbox, so that track has no result. But I don not find the reason why yolox don not work well? Help! problem3

soumajm commented 1 year ago

What kind of images are you testing on? The sample image you have here does not include any person.