zhaipro/easy12306

Implements a two-stage CNN pipeline with separate models for text recognition (model.h5) and image classification (12306.image.model.h5) to solve grid-based CAPTCHA challenges requiring object identification and localization. Uses pre-trained Keras models accepting JPG inputs and outputs grid coordinates with confidence scores for targeted objects. Earlier versions employed image similarity matching; current approach leverages deep learning for improved accuracy on unseen verification samples.

2,917 stars. No commits in the last 6 months.