I’m gonna add the missing stuff to the readme also while I’m at it.
It would probably be good to make the whole thing a bit more clear.
I’m gonna add this image that @MikeLud1 sent + one showing the recording tab with jpegs enabled since a few people have been tripped up there.
View attachment 213527
A common roadblocks / troubleshooting section would be good to have also detailing BI configuration stuff or AI issues.
Thank you for pointing that out. I thought you had to have the JPEG recording on for it to work? I ran into some sort of problem related to that early on, but maybe it was something else.I genuinely disagree with some of the settings being flagged as required:
- The To Cancel field can be left empty because, given how the custom model is configured (objects:0, alpr), this setting becomes redundant.
- Is using the main stream truly beneficial for you? In my experience, it only slows down CPAI without offering any accuracy improvements.
- I'm not saving JPEGs anywhere—neither in the record tab nor in the alerts list. While saving JPEGs can be required for other purposes or for personal taste, it is not necessary for ALPR.
Generally main steam is not needed for AI. For the ALPR module it needs the full resolution image for the cropped license plate OCRIs using the main stream truly beneficial for you? In my experience, it only slows down CPAI without offering any accuracy improvements.
Generally main steam is not needed for AI. For the ALPR module it needs the full resolution image for the cropped license plate OCR
If you use main stream it should improve the OCR.Hmm... I just double-checked, and I don't have the main stream enabled. However, I am using 720p as the sub-stream for my ALPR camera. Maybe that's why it works well for me then?
Is that what this checkbox does?Generally main steam is not needed for AI. For the ALPR module it needs the full resolution image for the cropped license plate OCR
Is that what this checkbox does?
View attachment 213614
so we can leave the sub stream enabled for N-up and previews, etc., but pass the main stream to AI for processing?
What about both left/right arrow keys and Next/BackRoadmap request: Add back and forward to image viewer @VideoDad
This was such a good idea. I just bound the left/right arrow keys instead of using buttons, but I think most would prefer that unless you're trying to use a tablet or something
View attachment 213629
Yeah it will be on :nightly instead of :latest.Is this feature available now?
chmod +x update.sh
I was thinking about it for tablet reasons bc I’m sure there are some homeassistanters with a wall mount that might want that.
I would just use the mouseDo u want it for touch screen reasons or just a mouse preference?
Once you do the &JSON switch I will add to the JSON the data to crop the characters so we can save OCR training imagesI’m gonna start the switch to &JSON in a bit. Expecting to finish it tonight.
Touch would be nice for my HA dashboard on an iPad.Do u want it for touch screen reasons or just a mouse preference?
LovelyOnce you do the &JSON switch I will add to the JSON the data to crop the characters so we can save OCR training images