Closed jonadis closed 1 year ago
Without logs it is difficult to track this down, usually that means the GPU is failing to keep up and frames are being delivered slowly. Can you send a picture of the system page while this is happening?
I bet it could be just that simple. I'm trying TensorRT and when things are fired up using TensorRT the GPU usage shows 97%-98% most of the time. I think I saw this same behavior when using OpenVino too but I'm not 100% sure. I'll switch back to OV and see if it improves.
oh, I see you are running detect at 30 fps on that camera, that is highly discouraged. The fps does not need to match the cameras fps
It's a doorbell that I can't adjust the output stream settings and I thought the detect FPS had to match. I'll change that, however that camera is not one of the ones that often has this symptom.
however that camera is not one of the ones that often has this symptom.
right, but it may be using many resources and frames that other cameras are trying to use
Understood, I'll lower them all down to 5fps and see how it goes.
worth noting that if you have an integrated GPU there is no reason one GPU has to be used for everything. For example I use my integrated GPU for the streams that I don't care about latency on and then use my nvidia GPU for the streams that I want the lowest detection latency possible
Good to know. I did fire things up with detect on 5FPS for all cameras and 2/3 of the cameras had the no frames message... I switched back to OpenVino and the issue seems gone so far. I do have a Coral but its a PCI Express one and this motherboard doesn't have enough PCI Express slots to hold both the GPU and the Coral, so I've planning to replace the motherboard.... Guess I'll fasttrack that now. Thanks for all your help.
this issue seems to have been tracked down and the behavior is expected given system page so will close, feel free to create a new issue if something else comes up
Describe the problem you are having
I'm really enjoying 0.13-beta1 however I've been seeing a lot of "no frames have been received, check error logs" on the main screen. When I click on an individual camera the feed fires up just fine, then going back to the main page sometimes that error clears some times it doesn't but there's never an issue with the camera feed, just the message on the main screen. This happens on various cameras, I'll include the FFProbe output below from one of them that's currently having issues but just about all of my cameras have done this from time to time. When looking at HomeAssistant I see the same "no frames have been received, check error logs" in the camera feed for a second or two and then it goes to the camera feed as expected. Log wise there's absolutely nothing in the log within 15+ minutes of my interaction with the web page. And that goes for the frigate logs, the go2rtc logs, and the nginx logs just have the typical "GET" messages in them.
Version
0.13.0-9185753
Frigate config file
Relevant log output
FFprobe output from your camera
Frigate stats
Operating system
Debian
Install method
Docker Compose
Coral version
CPU (no coral)
Network connection
Wired
Camera make and model
Mostly HikVision
Any other information that may be helpful
No response