Korean car license plate recognition using LPRNet

This repository is based on the paper LPRNet: License Plate Recognition via Deep Neural Networks.

We use the KarPlate Dataset for training and test model

Dependencies

• Python 3.6+
• Tensorflow 1.15 or 2
• Opencv 4
• tqdm
• editdistance

Usage

Pre-trained model

Download best_weights.zip, unzip and move into saved_models folder for testing

Demo

python predict.py -i data/test_images/4.jpg -w saved_models/weights_best.pb

Training

python train.py -l data/label.json -i data/train_images --valid_label data/test.json --valid_img_dir data/test_images --save_weights_only --load_all 

Testing

python predict.py -i data/test_images/4.jpg -w saved_models/weights_best.pb

Evaluate

python evaluate.py -l data/test.json -i data/test_images/ -w saved_models/weights_best.pb