- Reproduce weak-supervision training as mentioned in the paper https://arxiv.org/pdf/1904.01941.pdf
- Generate character bbox on all the popular data sets.
- Expose pre-trained models with command line interface to synthesize results on custom images
git clone https://github.com/autonise/CRAFT-Remade.git
cd CRAFT-Remade
conda env create -f environment.yml
conda activate craft
pip install -r requirements.txt
Put the images inside a folder.
Get a pre-trained model from the pre-trained model list (Currently only strong supervision using SYNTH-Text available)
Run the command -
python main.py train_synth --mode=synthesize --model=./model/final_model.pkl --folder=./input
SynthText - https://drive.google.com/open?id=1qnLM_iMnR1P_6OLoUoFtrReHe4bpFW3T
- ICDAR 2013 - In Progress
- ICDAR 2015 - In Progress
- ICDAR 2017 - yet_to_be_completed
- Total Text - yet_to_be_completed
- MS-COCO - yet_to_be_completed
- ICDAR 2013 - In Progress
- ICDAR 2015 - In Progress
- ICDAR 2017 - yet_to_be_completed
- Total Text - yet_to_be_completed
- MS-COCO - yet_to_be_completed
Download the pre-trained model on Synthetic dataset at https://drive.google.com/open?id=1qnLM_iMnR1P_6OLoUoFtrReHe4bpFW3T
Make your own custom dataloader as in train_weak_supervision/dataloader.DataLoaderMIX
Run the command -
python main.py weak_supervision --model=/path/to/pre-trained/Synth-Text-Model --iterations=epochs-of-weak-supervision