Kanav Dawra
With the advent of machine learning digital optical character recognition (OCR) is mostly solved and not much research is required in solving that problem. But Handwriting recognition is still an ongoing research topic because of many challenges. My hope is that by using a rich dataset of handwritten words and using deep learning techniques I can contribute in some way
The challenge with this project was to get a rich enough and small enough dataset that can be trained on my limited hardware. Another challenge was to integrate the web app into my website without breaking functionality.
Build a CRNN model which can successfully recognise handwriting with reasonable accuracy and ability to be integrated into any website
Dataset used in this project is the transcriptions of 400,000 handwritten names on Kaggle. This dataset contains 400k+ names since these are names that mean there are potential 400K handwritings.
Right now it can only predict a sentence that is 35 characters long but in future, there is potential for making it recognize a full-page article