- Python 2.7
- Tensorflow 1.2.0
- COCO API
- Use PubLayNet dataset. Download
labels.tar.gz
and decompress it. - Run
python preprocess_doc.py
for preprocessing dataset. - Run
bash ./experiments/scripts/train_doc.sh
to train a model.
- Use Rico dataset. Download
semantic_annotations.zip
and decompress it. - Run
python preprocess_mobile.py
for preprocessing dataset. - Run
bash ./experiments/scripts/train_mobile.sh
to train a model.
For more details, please see the original README.