ajeet28/EAST.pytorch

A PyTorch implementation of EAST: An Efficient and Accurate Scene Text Detector

EAST Text Detection Model in PyTorch

- with single QUAD Representation of Bounding Boxes

A PyTorch implementation of EAST: An Efficient and Accurate Scene Text Detector for bounding box detection

References:

1. https://arxiv.org/pdf/1704.03155.pdf
2. https://github.com/liushuchun/EAST.pytorch
3. https://www.pyimagesearch.com/2018/08/20/opencv-text-detection-east-text-detector/

Steps:

1. Data folder should be organised as follows:
The data dir should have 3 sub dirs namely, "train", "dev" and "test".
The "train" and "dev" dirs should have 2 sub dirs namely "images" and "annotations"
The "test" dir should have a sub dir "images"

2. Run format_data.py

3. Train the model using train.py

4. Evaluate the model using eval.py

5. Predict using test.py