CRNN_Project

CRNN PROJECT

TODO

• Completed training on 3.6 million data to achieve 96% accuracy
• Provide training interface to individual datasets
• Merging CTPN for complete OCR recognition
• Provide ID card identification interface
• Provide business licence identification interface

Dev Environments

1. WIN 10 or Ubuntu 16.04
2. PyTorch
3. tensorboardX
4. opencv

Data

Synthetic Chinese String Dataset

1. Download the dataset
2. Edit Dataset.data.dataset.py self.root to you image path
3. Download the labels (password: eaqb)
4. Put char_std_5990.txt in dataset/data/text
5. And put train.txt and test.txt in dataset/data/text

Demo

Note: The ID card numbers appearing in the pictures above are self-made ID card data set numbers (not real resident ID card numbers).

Reference

• https://github.com/meijieru/crnn.pytorch
• https://github.com/Sierkinhane/CRNN_Chinese_Characters_Rec
• https://github.com/PaddlePaddle/PaddleOCR