/torch_backbones_reference

Unofficial implementations of some classical CNN backbones with pytorch

Primary LanguagePython

torch_backbones

  • Unofficial implementations of some classical backbone CNNs with pytorch

  • This is pretty much just a practising project. If you want to do some research, I strongly recommend using officially issued models in torchvision.

  • Requirements：

    · torch，torch-vision

    · torchsummary

    · mlflow

    · onnx onnnruntime

    · openvino

  • A training script is supplied in “train.py”，the arguments are in “utils/args.yml".

  • new features：

    1. mlflow traning track. You need a mlflow server, and to intall mlflow on you training server.
    2. onnx inference.
    3. openvino inference.

    N.B. If you don't want these features, pelease check release branch v1.0.0.