Skip to content

Latest commit

 

History

History
55 lines (35 loc) · 2.68 KB

readme.md

File metadata and controls

55 lines (35 loc) · 2.68 KB
Raw

TRIE

This code repository contains the implementations of the paper TRIE: End-to-End Text Reading and Information Extraction for Document Understanding (ACM MM20).

Preparing Dataset

  • SROIE: Dataset used for ICDAR2019-SROIE competition. Since this dataset contains many annotation errors, we will clean this dataset later and release corresponding trained model soon.
  • Wildreceipt: Original dataset can be downloaded from MMOCR. Formatted training datalist and vocab used can be found in demo/text_ie/datalist.

Train From Scratch

Temporarily, we release pretrained model on Wildreceipt dataset using detection and recognition ground-truth, more trained models on diverse datasets will be published soon.

If you want to re-implement the model's performance from scratch, please following these steps:

1.Firstly, download original dataset and unzip it to demo/text_ie/datalist.

2.Secondly, modify the dataset, word dir and pretrained resnet50 model path in demo/text_ie/trie/config/wildreceipt_gt_trie.py.

3.Thirdly, direct run demo/text_ie/trie/dist_train.sh.

We provide the implementation of online validation

Test

Given the trained model, direct run demo/text_ie/trie/test.sh to inference model.

Trained Model Download

All of the models are re-implemented and well trained based on the opensourced framework mmdetection. So, the results might be slightly different from reported results.

Results on various datasets and trained models can be download as follows:

Dataset Backbone Pretrained Train Scale Test Scale F1-score Links
Wildreceipt ResNet50 ImageNet (512, 512) (512, 512) 87.08 config, pth (Access Code:6xq7)

Citation

If you find this repository is helpful to your research, please feel free to cite us:

@inproceedings{zhang2020acmmm20,
  title={TRIE: End-to-End Text Reading and Information Extraction for Document Understanding},
  author={Peng, Zhang and Yunlu, Xu and Zhanzhan, Cheng and Shiliang, Pu and Jing, Lu and Liang, Qiao, and Yi, Niu and Fei, Wu},
  booktitle={Proceedings of the 28th ACM International Conference on Multimedia (ACM MM)},
  pages={1413–1422},
  year={2020}
}

License

This project is released under the Apache 2.0 license

Contact

If there is any suggestion and problem, please feel free to contact the author with [email protected].