marcoslucianops / DeepStream-Yolo

NVIDIA DeepStream SDK 7.0 / 6.4 / 6.3 / 6.2 / 6.1.1 / 6.1 / 6.0.1 / 6.0 / 5.1 implementation for YOLO models
MIT License
1.38k stars 343 forks source link

Weaker detections in deepstream-app #520

Open chandana-21 opened 3 months ago

chandana-21 commented 3 months ago

I ran inference on the same video using my yolov5l .pt weights using detect.py from yolov5 github repo and onnx weights in deepstream.

This is the detect.py video

https://github.com/marcoslucianops/DeepStream-Yolo/assets/79861528/cfd7129e-d9f6-49d2-860c-0a7431830888

This is the video in deepstream with config:

nms-iou-threshold=0.45
pre-cluster-threshold=0.25
topk=300

https://github.com/marcoslucianops/DeepStream-Yolo/assets/79861528/d07aa450-15be-477b-8082-e92ec8303b1f

and this is the video with config:

nms-iou-threshold=0.65
pre-cluster-threshold=0.001
topk=300

https://github.com/marcoslucianops/DeepStream-Yolo/assets/79861528/e063b8e5-0236-43a3-a9b0-1227b19dda20

The detections are missing for most of the second half of the video in deepstream, what could be causing this and how can i solve it?