More details

[folkien]

Could you explain better the idea behind this projected? As far I am concerned, it looks like you :

1. Trained YOLO to recognize (Letters/Numbers + License Plate)
2. Parse video, detect each frame
3. Filter only detections inside 'License Plate detection bbox'
4. Read detections labels from left to right

[stephanecharette]

Correct. The process I was showing with this project is to use the same YOLO network twice. Once to find the plate, then again to read the characters. No need to have 2 networks, or use tesseract to "read" the plates.

[stephanecharette]

This was part of a long discussion that took place on the Darknet/YOLO discord, come join if you want more details: https://discord.gg/zSq8rtW