Have you tried this. Convolutional recurrent network in pytorch. https://github.com/meijieru/crnn.pytorch https://github.com/meijieru/crnn.pytorch I think this

Have you tried this. Convolutional recurrent network in pytorch. https://github.com/meijieru/crnn.pytorch https://github.com/meijieru/crnn.pytorch I think this is the current state-of-the-art (or close to it) for OCR on noisy low-res images. Here is a broader list of resources/papers for scene text recognition https://github.com/chongyangtao/Awesome-Scene-Text-Recognition https://github.com/chongyangtao/Awesome-Scene-Text-Recogniti...