transport prediction to cpu lost too much time

Autonomous Machines Jetson & Embedded Systems Jetson AGX Xavier
1

My prediction is segmentation result with size(976x208), when transport result to cpu, it could lost 60ms in xavier. My tensorrt version is 5.0.2.6, is there some way to solve this problem? Thank you very much.

2

Hi mengzhibin321,

Please try with highest performance setting to see if get improved.

sudo nvpmodel -m 0
sudo jetson_clocks