DeepStream-Yolo icon indicating copy to clipboard operation
DeepStream-Yolo copied to clipboard

Published 20 hours ago • verified publisher icon marcoslucianops

yolov5 and deepstream accuracy problem

Open qybing opened this issue 2 years ago • 3 comments

The tensorrt weights I exported using the yolov5 official website are almost the same as the original weight prediction results, and the error range is also within an acceptable range, but the weights I derived according to your method are sometimes very different in the results predicted by Deepstream. Do you have any suggestions?

qybing avatar Aug 25 '22 01:08 qybing

I'm trying to increase the accuracy of YOLOv5 on DeepStream. The current implementation is the most accurate I've gotten so far. The TensorRT layers (from DeepStream) are the same PyTorch layers, but there's an accuracy drop on DeepStream.

marcoslucianops avatar Aug 25 '22 22:08 marcoslucianops

I am also getting really Low Accuracy on yolov5. Is this relates to waning I get while converting model "Warning [TRT: Subnormal FP16 values detected." ?

Arpit1496 avatar Sep 07 '22 20:09 Arpit1496

I am also getting really Low Accuracy on yolov5. Is this relates to waning I get while converting model "Warning [TRT: Subnormal FP16 values detected." ?

It's not related to that.

marcoslucianops avatar Sep 23 '22 13:09 marcoslucianops

okay. Any updates on this ? How can we help to improve the accuracy ?

Arpit1496 avatar Dec 08 '22 19:12 Arpit1496

No updates yet.

marcoslucianops avatar Jan 26 '23 10:01 marcoslucianops

@marcoslucianops Does this happens as well in recent versions of YOLO?

Alberto1404 avatar Jan 30 '23 16:01 Alberto1404

@marcoslucianops Does this happens as well in recent versions of YOLO?

Yes, but, in my use cases, it's not a relevant drop, so I still using DeepStream and YOLO on projects.

marcoslucianops avatar Jan 31 '23 03:01 marcoslucianops

I've noticed this for yoloV8 also. Are there any updates?

OctaM avatar Mar 21 '23 13:03 OctaM

I've noticed this for yolov7-tiny. I have a model for 6-classes, and one particular class has a major drop in comparisson with pytorch

sruap1214 avatar Apr 04 '23 22:04 sruap1214

Hi @marcoslucianops we're having this issue too but are keen to help resolve it. Do you have any suggestions on where we should start?

We have also attempted to resolve this by using yolov8 as an onnx model, which had similar results.

We're seeing a significant drop in accuracy for some objects (40%) when running inference with ultralytics yolov8 vs in deep stream. With yolov4 we do not see much of a drop in accuracy comparison to running inference with darknet vs deep stream.

Any tips or comments on where we should begin would be greatly appreciated! Thanks!

pullmyleg avatar Apr 27 '23 02:04 pullmyleg

Is there a certain YOLO model that works fine after deployment without any issues?

KindOfBlue7 avatar May 09 '23 06:05 KindOfBlue7

We have also attempted to resolve this by using yolov8 as an onnx model, which had similar results.

https://github.com/marcoslucianops/DeepStream-Yolo/issues/339

Is there a certain YOLO model that works fine after deployment without any issues?

I'm adding new benchmarks to the README.md. You will be able to see the results and choose the best option.

marcoslucianops avatar May 19 '23 12:05 marcoslucianops