Jiarun Liu

Hi! Thanks for your great work. The onnx_runtime_gpu version works fine to me, and I'm trying to deploy a faster TensorRT version. I've followed the instructions of [onnxruntime_tensorrt.md](https://github.com/xmba15/onnx_runtime_cpp/blob/master/docs/onnxruntime_tensorrt.md) to build...

Performance on novel objects

1

Thanks for sharing the great work. I wonder how does the model works for novel objects that are not seen during training phase. Have you ever conducted similar experiments？