What resolution did you use on benchmarking

I’m try to use yolov3-tiny on jetson nano.

What resolution did you use in https://developer.nvidia.com/embedded/jetson-nano-dl-inference-benchmarks?

If this benchmark use HD video, how long does it take at full HD?


Hi cansee5744, the input resolution to Tiny-YOLOv3 network is 416x416. The images/video get downsampled to that, so the network processing time is somewhat independent of the original image’s resolution.

See here for the directions for obtaining and running the benchmark: https://devtalk.nvidia.com/default/topic/1050377/jetson-nano/deep-learning-inference-benchmarking-instructions/

Note that some of the images from the provided test set are full HD (1920x1080), others are 1280x720. You can substitute whichever images you would like however.