It started up fine, I cant tell a difference yet. I did look at debug again before I killed it and It was back down to 3.2. Right now it is showing 10. Ill report back once it gets busier outside and they start doing some work. One thing I noticed though, it use to always have a box around my car when it is in the garage and its not doing that now…I did walk in front of the driveway cam and it boxed me:)
da@deep:/opt/frigate$ nvidia-smi
Thu Apr 9 10:11:15 2020
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 440.64.00 Driver Version: 440.64.00 CUDA Version: 10.2 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 GeForce RTX 2080 On | 00000000:07:00.0 Off | N/A |
| 24% 25C P8 4W / 245W | 749MiB / 7973MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| 0 1620 G /usr/lib/xorg/Xorg 38MiB |
| 0 1880 G /usr/bin/gnome-shell 17MiB |
| 0 32495 C python3 681MiB |
+-----------------------------------------------------------------------------+
Car just drove by that really goes through both drive and front cams…i quickly checked debug and it was at 2.49…maybe it reports 10 when idle? wild ass guess
It starts at 10 as a default value and computes a rolling average. If your speeds are less it will need to run inference several times to come down to the actual time.
So we had an intruder in our back yard last night, so I’ve rolled back to 0.2-beta because I need it to stay up longer than a few days… I will try testing again when there are potential fixes for the fmpeg_process didnt return any bytes issue.
Hey! So it looks like you’ve done a ton of work in the last year! I recently updating the docker package to the latest and well, it’s completely broken now. Is there any way you could add any additional information needed to migrate?
I read your github, updated my config to match your latest config type. But I don’t know if there are more steps? Do I need to add ffmpeg to my directory now? Do you have a changelog somewhere?
Here is my current error
File “detect_objects.py”, line 79, in main
cameras[name] = Camera(name, FFMPEG_DEFAULT_CONFIG, GLOBAL_OBJECT_CONFIG, config, prepped_frame_queue, client, MQTT_TOPIC_PREFIX)
On connect called
File “/opt/frigate/frigate/video.py”, line 131, in init
self.regions = self.config[‘regions’]
KeyError: ‘regions’
error about regions suggest that you are not using latest version of code?
as it does not use regions anymore.
i dont have any knowledge about unraid, but make sure that you are running latest release.
latest release image shoud be blakeblackshear/frigate:0.5.0, or prerelease images 0.5.1-rc2.
I went back to .4 beta and things and been running flawlessly.I think the new versions with the auto regions are going to be awesome once a bit more stable. For me I already had the regions setup and its been rock solid.
I am testing out the frigate-gpu image and have it working amazingly well on a GTX1060.
My cams produce 2x streams (both 20fps)…a low-res h.264 (640x360) and full res (6Mp) H.265…this works for both!
The problem is things are being missed on the low res stream, so I thought I could use the full res stream and just change the resolution within the config.yml.
However when I try to drop the height + width of the stream in the config.yml the stream starts but I just get static on the debug stream and nothing is detected.
Also - is there a way of reducing the running CPU overhead - its causing my i5 8400 to run at 20 - 30% constantly on the h.265 stream and 15 - 20% on the h.264 stream. I currently have the FPS set to 5 for 2 cameras and the ‘take_frame’ to 1.
The GPU running at ~20% when persons are detected which is totally acceptable.
