pytorch tensor and inference

[swkms]

Hello, I use VPF decoding and yolo model, nv12 ->yuv420 ->rgb24 ->rgb24Planar ->python tensor ->"run_inference" in the example. Can I use rgb24Planar to give self. bindings [0] without converting to python tensor

[contentis]

If you want to use the python interface, you need to use the tensor conversion as the binding is a raw byte stream that needs to be interpreted. This should not add any overhead and therefore is not an issue.