Toybrick

标题: time spent difference between rknn.inference and rknn.eval_perf [打印本页]

作者: pangu    时间: 2019-3-15 16:48
标题: time spent difference between rknn.inference and rknn.eval_perf
I ran 100 times of  rknn.inference() of mobilenet-v1-ssd and measure average time of each run using time.time(). It is

Time:  0.0248 sec

However, rknn.eval_perf() shows:

========================================================================
Total Time(us): 13779 (0.013779 sec)
FPS: 72.57
========================================================================

Both has the same input image, How can rknn_eval time be almost half of rknn.inference? Which one is more accurate?

Thanks,


作者: 程子    时间: 2019-3-15 17:01
好像是常规现象。
我自己网络 rknn.eval_perf() 返回的时间大概是 rknn.inference() 的70%
作者: tof3d    时间: 2019-3-15 17:13
I have the same question
作者: zhangzj    时间: 2019-3-18 16:04
rknn.inference()比rknn.eval_perf()耗时久是正常的。
inference()包含了数据通信拷贝等一系列额外的开销。
eval_perf()是用作性能评估,仅统计网络推理运行时间。  




欢迎光临 Toybrick (http://t.rock-chips.com/) Powered by Discuz! X3.3