Skip to content

Latest commit

 

History

History
46 lines (25 loc) · 978 Bytes

README.md

File metadata and controls

46 lines (25 loc) · 978 Bytes
Raw

ICDAR2015 😄

  • This repo is about to implement diffrent kinds of modern toolkits to detect and recognize texts in image. In the repo so far, I used DB and CRNN (implemented in mmocr) and also, I tried a simple CRNN model written in Keras for recognizing text, which is much depended on this

  • Let's start with:

git clone https://github.com/manhph2211/ICDAR-2015.git 
cd ICDAR-2015
bash bash.sh

Datasets

  • This implement uses datasets from ICDAR-2015 competiton. You can download data from the offical website (in the task 4.1 & 4.3). All zip files should be unzipped at /data

Text Localization

Training (db-mmocr)

cd mmocr
python3 ./tools/train.py ../detector.py

Text Recognition

Training

  • With crnn-mmocr:
cd mmocr
python3 ./tools/train.py ../recognizer.py
  • With simple CRNN(Keras), just run python3 train.py