YCG09/chinese_ocr
A deep learning system for Chinese text detection and recognition in images using CTPN and DenseNet with CTC loss.
Not currently ranked — collecting fresh signals.
star history
Implements end-to-end Chinese OCR with two stages: CTPN (Connectionist Text Proposal Network) for detecting text regions and a DenseNet encoder with CTC (Connectionist Temporal Classification) decoder for sequence recognition. Built on Tensorflow and Keras with complete training scripts, synthetic data generation support, and demo evaluation pipeline. The model achieves 98.3% validation accuracy at 8ms inference time.
Frequently asked
- What is YCG09/chinese_ocr?
- A deep learning system for Chinese text detection and recognition in images using CTPN and DenseNet with CTC loss.
- Is chinese_ocr open source?
- Yes — YCG09/chinese_ocr is open source, released under the Apache-2.0 license.
- What language is chinese_ocr written in?
- YCG09/chinese_ocr is primarily written in Python.
- How popular is chinese_ocr?
- YCG09/chinese_ocr has 2.8k stars on GitHub.
- Where can I find chinese_ocr?
- YCG09/chinese_ocr is on GitHub at https://github.com/YCG09/chinese_ocr.