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 344 forks source link

Yolov7 - Segmentation fault #356

Open sruap1214 opened 1 year ago

sruap1214 commented 1 year ago

Hi Marcos

I have been using your repository for several months. Recently I update the compilation with the last version of the repo, but when I executed my code in the container it says segmentation fault for Yolov7. I go back to commit "ab6de54c4398c1daeb6fc31b06af29f97663f211" and works perfectly. Nevertheless with the last commit happen that

sruap1214 commented 1 year ago

I double check and the commit that generates the problem is the one called "Big update" 07feae9509ffb581fa85f65bec653d1c1d001056.

I tried the commit before that one and works perfectly for yolov7. I am trying in with docker container in a x86 and discrete GPU (RTX 3080)

marcoslucianops commented 1 year ago

The new update only work with engines exported with the new method. You need to compile the new lib and use the export_yoloV7.py to export your model to ONNX and run with the config_infer_primary_yoloV7.txt file with the exported model on DeepStream.