JuZi233

cpu consumption is a bit high, I hope it can be optimized

3


yolov8.hpp下的： nvinfer1::ICudaEngine* engine = nullptr; engine的大量方法例如：getBindingDataType，getBindingName，destroy等在tensorrt10中已移除

yolov8.hpp下的： nvinfer1::ICudaEngine* engine = nullptr; engine的大量方法例如：getBindingDataType，getBindingName，destroy等在tensorrt10中已移除

Can the dll of the cuda version of opencv generated on 3080 (compute-> 8.6) be used on 40 series N cards (compute->8.9)? What about the 20 series and 50 series?

I suggest removing the configuration condition of cuda12.8. Currently, I just want to build the dml version. It cannot be generated normally without installing cuda.