Hi, I’m currently using opencv to perform inference on images using my trained model exported to onnx. It works pretty well and I take advantage of the stock, OpenVino and CUDA backends.
My biggest complaint is the size of all the dlls that needs to be redistributed for CUDA/cuddn support but that’s nothing OpenCV teams can improve.
While asking around, I heard about the alternative (at least for Windows) called DirectML.
Are there any plans to support it has a backend? I’ve looked at the source code and it seems that G-API seems to support onnx inference using DirectML. Is G-API meant to replace dnn in a near future? Where can I find information about that?
Thanks in advance