train with my own datasets, the results is unstable

[MZener]

Hi,
Thanks for your sharing.
I train with my own datasets. The results are as below：
Two consecutive images are extracted from a video stream. the detected line is unstable, What reason is this?
Thanks a lot.

[cherubicXN]

Yes, it is indeed a problem. Another example can be found in a Youtube video. The main reason for this problem is that we did not consider the temporal information in both our designs and training.

[yizhidecainiao]

Hi, Thanks for your sharing. I train with my own datasets. The results are as below： Two consecutive images are extracted from a video stream. the detected line is unstable, What reason is this? Thanks a lot.

Hi, how do you train your own data, can you share your configuration file and the training commands, thanks a lot!