I converted my custom yolov3 model to onnx then onnx to tesnsorrt model on Jetson nano, it is taking 0.2 sec to predict on images i tried it on video and it is giving only 3 fps is there any way to increase this.
1. Please maximize the device performance with the following command first:
$ sudo nvpmodel -m 0 $ sudo jetson_clocks
2. You can run the YOLOv3 model on FP16 mode for a better performance.
$ /usr/src/tensorrt/bin/trtexec --onnx=/usr/src/tensorrt/data/mnist/mnist.onnx --fp16