The source codes has been released, will make it clear for those who are not familiar with deep learning and encoder-decoder models:
The data will be released after it is prepared:
- Tree Decoder: A Tree-Structured Decoder for Image-to-Markup Generation
Preprocessing of training set. (.pkl)
python data/gen_pkl.py --dataset_type CROHME --op_mode train
Preprocessing of test set. (.pkl)
python data/gen_pkl.py --dataset_type CROHME --op_mode test
python data/gen_voc.py --dataset_type CROHME
python codes/latex2gtd --dataset_type CROHME
python codes/prepare_label.py --dataset_type CROHME
python codes/train_wap.py --dataset_type CROHME
python codes/translate.py --dataset_type CROHME --batch_size 8 --K 112 --k 3 --model_path ../train/models/210418/WAP_params_last.pkl --dictionary_target ../data/CROHME/dictionary.txt --dictionary_retarget ../data/CROHME/relation_dictionary.txt --fea ../data/CROHME/image/offline-test.pkl --output_path ../test/