inference time about lanenet-lane-detection HOT 6 CLOSED

@klnsz Most time may be caused during the mean shift clustering process

from lanenet-lane-detection.

@klnsz Most time may be caused during the mean shift clustering process

Thank you for your amazing work ,however ,when i test the pretrain model on my dataset, i found the preprocess time is around 0.15s , postprocess time is around 0.2s, the most time consume is the network predict time, which is up to 3s,
emm...,may you tell me your test time and your GPU type?

from lanenet-lane-detection.

@ktr-hubrt Usually most time should be cost during post processing time which includes a mean shift cluster procedure. I think you may check if you use the cpu to apply inference procedure. By the way I use gtx 1070 on my computer.

from lanenet-lane-detection.

作者你好，我发现把一张图片放在不同batch中验证，都到的测试结果都不同？，网络的bn有问题？

from lanenet-lane-detection.

作者你好，我发现把一张图片放在不同batch中验证，都到的测试结果都不同？，网络的bn有问题？

@MaybeShewill-CV

from lanenet-lane-detection.

Hi @klnsz, could you please let me know if anything special is needed to install this on TX2? Which version of L4T and cuda you are using on Tx2?

from lanenet-lane-detection.