search

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
1.39k stars 344 forks source link

Is support yolov4-p6? #85

Closed amelieko closed 2 years ago

amelieko commented 2 years ago

Hi, @marcoslucianops, I use Yolov4-p6 with DeepStream 5.1 . I follow the step and use DeepStream-Yolo/native to generate engine file. I changed the config_infer_primary.txt and deepstream_app_config.txt to match my setting. However, after execute , I got error message as below,

deepstream-app: yolo.cpp:361: NvDsInferStatus Yolo::buildYoloNetwork(std::vector&, nvinfer1::INetworkDefinition&): Assertion `0' failed. Aborted (core dumped)

I have no idea whether the .pt file is not right or wrong setting. Or, yolov6-p6 is not supported now? Please give me some advise. Thanks a lot,

marcoslucianops commented 2 years ago

You did something wrong, here it works normally.

marcoslucianops commented 2 years ago

yolov4-p6.cfg yolov4-p6.weights

amelieko commented 2 years ago

Hi, @marcoslucianops ,

I execute this setting and it's work. Thanks for helping to check, In my situation is, I use Pytorch-Scaled-YOLOv4 to build new model. It's work when I do inference use python. However, when I generate engine. it still show error message.

Loading pre-trained weights deepstream-app: utils.cpp:90: std::vector loadWeights(std::__cxx11::string, const string&): Assertion `file.gcount() == 4' failed. Aborted (core dumped)

maybe the config setting is wrong. I have two classes and I modified the cfg. file for four yolo block,( changed parameters: classes=2, filter=24, activation=linear) Did I miss other settings? Thanks again.

amelie

marcoslucianops commented 2 years ago

Are you using the Darknet model or PyTorch model? The PyTorch model isn't supported for now.

amelieko commented 2 years ago

Ohhhh....That's why I can't execute my model because I use Pytorch model. anyway, thank you very much.