added
- The new yolov8 and yolov11 model is trained with bbox and the key 4 points of the license plate instead of the yolo+craft
- Upscaling model for license plates with low resolution https://github.com/ria-com/upscaler/tree/main
updates
- Update pipelines (see https://github.com/ria-com/nomeroff-net/tree/v4.0/examples/ju/inference for more)
bugfix
- Repair bug with different pytorch_lightning versions for ocr models
deleted
- yolov5 support
- yolovx support
- unused experimental orientation number plate classificator
- unused experimental inverse number plate classificator
- unused experimental fraud number plate classificator