search

NVIDIA-AI-IOT / yolo_deepstream

yolo model qat and deploy with deepstream&tensorrt
Apache License 2.0
533 stars 135 forks source link

batch inference fps calculation #40

Closed victorzhu-epirus closed 6 months ago

victorzhu-epirus commented 1 year ago

just wondering how the fps is calculated for the batch 16 case. It reported https://github.com/NVIDIA-AI-IOT/yolo_deepstream/tree/main/yolov7_qat the fps is 168 for FP16. Does this mean for each batch of 16 it will take 168/16=10ms?

wanghr323 commented 1 year ago

fps : frame per second, fps is 168 for FP16. that means one image will take 1/168 s, one batch will take 16/168 s.

wanghr323 commented 6 months ago

close as no feedback for long time