@ZhaiSoul commented on GitHub (Feb 4, 2026):
Would you consider adding motion-based classification? That is, sending the parts of the frame with movement to the classification model.
@bench66517 commented on GitHub (May 23, 2024):
I have the same problem. Vlc is easy able to connect to the stream. But with frigate i‘m not able to connect because it takes a long time…
@NickM-27 commented on GitHub (Feb 4, 2026):
we could look at that in the future, though that has a lot of ramifications that are not currently present.
@villasenor commented on GitHub (May 26, 2024):
I've seen this with IMOU (off-brand Dahua) cameras over the years sporadically, even when the cameras are on ethernet (although it happens less…
@coldkevin commented on GitHub (Jun 4, 2024):
Im having the exact issue as described by @villasenor . The stream at some point has an issue which causes the "no frames detected" sceen in frigate…
@aburow commented on GitHub (Aug 21, 2024):
I have the same issues with timeouts on Dahua cams. Can't create or edit regions on Dahua cameras either. There was some improvement in the last…
@github-actions[bot] commented on GitHub (Sep 21, 2024):
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs.…
@raidenlal commented on GitHub (Apr 19, 2024):
It seems that during night lower bitrate somehow delays first complete frame. Dont know why. Camera is Xiao Fang.
@NickM-27 commented on GitHub (Feb 4, 2026):
This is likely best run using a classification model or AI directly
@github-actions[bot] commented on GitHub (May 20, 2024):
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs.…
@ZhaiSoul commented on GitHub (Feb 4, 2026):
This is likely best run using a classification model or AI directly
For classification, collecting a dataset can be quite challenging, and you…
@NickM-27 commented on GitHub (Feb 4, 2026):
Within Frigate, but it could fairly simply be done via automation which in general is my point, using object detection has the same training problems…
@raidenlal commented on GitHub (May 20, 2024):
Still relevant. Frigate is unable to fetch image from camera almost everyday - while VLC is able to.
@NickM-27 commented on GitHub (Apr 19, 2024):
why does it take so long to start the stream? if the camera has an iframe interval that far apart that is going to cause many other problems
@NickM-27 commented on GitHub (Apr 22, 2024):
I don't think this will make sense. With the new UI coming in the next version of frigate more things are going to be configurable in the UI, and…
@NickM-27 commented on GitHub (Apr 23, 2024):
the default model is based off of the COCO dataset, additional labels can not be added to the default model. Frigate+ requests should go to the…