Frigate+ yolov9 models don't work with LPR as written in the documentation. #20205
-
|
Using any Frigate+ yolov9 model with the settings shown below, if you add license_plate under tracked objects it immediately breaks the LPR. The only way to resolve it and get it to work again is to remove it. This is directly contrary to the documentation: "Model Requirements i.e following the official documentation will directly prevent LPR from working in Frigate+. |
Beta Was this translation helpful? Give feedback.
Replies: 2 comments 7 replies
-
|
The documentation states that when running a Frigate+ model (or any model that natively detects license plates), you should add license_plate to the list of objects to track to improve LPR accuracy and performance(1)(2). However, you are experiencing a situation where including license_plate under track causes LPR to stop working, and removing it restores functionality. There is no information in the official documentation indicating that this behavior should occur or that adding license_plate to the track list would break LPR(1)(2). The documented guidance is that including license_plate as a tracked object is required for Frigate+ models to enable native license plate detection and associated LPR features(1)(2). If following the documented configuration prevents LPR from working with Frigate+ YOLOv9 models, this is not reflected in the official documentation(1)(2). You may want to check for updates or clarifications in the documentation or Frigate+ model documentation and LPR documentation. 📚 Sources: Have another question? Just tag @inkeep. |
Beta Was this translation helpful? Give feedback.
-
|
All of the Frigate maintainers are running Frigate+ models with LPR setup, and it is working correctly. This also includes many known users who have tested LPR in both standard and dedicated-LPR mode. More information is needed to understand why you think it is not working correctly, difference in behavior, etc. |
Beta Was this translation helpful? Give feedback.
-
|
I appreciate the help. Per the image below I've got 2000+ license plates tracked on a dedicated lpr camera that has worked flawlessly for a couple of weeks: 
That is with this config: Since I was so pleased (with this and other things, I have 9x2k cameras), I purchased Frigate+. However, according to the official documentation with Frigate+ you need to track "license_plates" as an official object. When I do this, in line with the documentation exactly - it no longer works. No plates show up in explore anymore. I'll check how do enable debug logs for LPR and see if I can see anything there. |
Beta Was this translation helpful? Give feedback.
-
|
lpr: LicensePlateRecognition: logger: Debug enabled, license_plate added under track. Logs: /api/Entrance1/latest.webp?height=240&cache=1758774766980 HTTP/1.1" 200 16342 "http://100.98.243.90:5000/" "Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:143.0) Gecko/20100101 Firefox/143.0" "-" According to Chatgpt everything is working based on these logs, yet as mentioned the second you include license_plates as a tracked object license plates stop apearing in "explore" tab. |
Beta Was this translation helpful? Give feedback.
-
|
It actually just caught two plates with the above settings. Now I think I see what's happening, before with the Yolox model it was capturing every plate without fail, but now with the Frigate+ Yolov9s model it's only capturing some of them. Are there any known config changes to fix this when moving to Frigate+ models? |
Beta Was this translation helpful? Give feedback.
-
|
After dropping the thresholds in Global LPR settings I've got it working correctly. p.s absolutely love Frigate and am very much looking forward to new features, especially related to face detection functionality! |
Beta Was this translation helpful? Give feedback.
After dropping the thresholds in Global LPR settings I've got it working correctly.
p.s absolutely love Frigate and am very much looking forward to new features, especially related to face detection functionality!