Skip to content

Latest commit

 

History

History
15 lines (9 loc) · 843 Bytes

README.md

File metadata and controls

15 lines (9 loc) · 843 Bytes

Hindi-On-Web

Hindi character recognition on web, with a drawing pad to sketch the character

Hindi Character Recognition on Sketchpad Demo

We used CNN's and streamlit to create a web app to recognize handwritten hindi characters.

Currently the network performs like trash :') but we are working on implementing transfer learning and working on a better dataset. The current model architecture is 3 conv blocks, leading into a fully connected layer to about 60 classes.

Future goals include implementing NLP to add dictionary support for full words and maybe sentences.

Dataset used for training model here

Massive thanks to Sree Harsha Nelaturu for helping with getting started on the streamlit front end