When using zed-mini camera + pytorch for target detection, an error occurred when the zed image was used as the input image for prediction, cuDNN error: CUDNN_STATUS_MAPPING_ERROR, I checked this seems to be caused by inconsistent context management of cuda, do you have a solution Way?
This error may also come from an unsupported GPU or software version mismatch. Are the ZED SDK and Pytorch (using CUDA) working fine independently?