A web app for annotating the text in images using state-of-the-art ML models.
This project is an attempt to combine Clova AI's implementations of text spotting (CRAFT: Github, Paper) and the best performing model in their text recognition benchmark (TPS-Resnet-BiLSTM-Attn: Github, Paper) in one simple web app that showcases an end-to-end implementation of the two. The models are both pretrained, using classes from the linked Github repos so the majority of code is taken from there and simplified down.
Install necessary packages
$ pip install -r requirements.txt
And run with
$ python app.py
The server should go up on localhost:5000
Simply open the app on localhost:5000 and browse to your image file. Click 'Submit' and after a few seconds your image should appear with annotations. All images are saved in static/output/
Before:
After:
