Thank you, and currently I’m still publishing scientific articles on this project and another one on NLP (it will be my next post). In fact, for this context of HTR, I only saw works improving the recognition of paragraphs/pages of text, but I’m not completely familiar with this subject. For digit recognition (HDSR), I believe you can use this project as well. For printed text there should be more options for optical models (unfortunately I’m not aware of this research area either), but you can also use this project as a way to validate the model.