I'm currently working on implementing YOLOv5 for object detection using GPUs. However, I've noticed that my system is consuming significantly more CPU resources than GPU resources during inference, which is not what I expected. I believe I have properly set up the environment and dependencies for GPU acceleration, including CUDA and cuDNN.
Despite these preparations, when I monitor system resource usage during inference, I'm observing higher CPU utilization than GPU utilization. This seems counterintuitive, as YOLOv5 is known for its GPU acceleration capabilities.
Are there any specific parameters or configurations I might be missing that could lead to this discrepancy in resource usage? Any insights or guidance would be greatly appreciated.
