Are you sure you want to delete this task? Once this task is deleted, it cannot be recovered.
wuzhf9 b902ff4e78 | 5 months ago | |
---|---|---|
.github | 8 months ago | |
configs | 6 months ago | |
deploy | 8 months ago | |
docs | 7 months ago | |
mindocr | 6 months ago | |
requirements | 8 months ago | |
tests | 7 months ago | |
tools | 7 months ago | |
.flake8 | 10 months ago | |
.gitignore | 9 months ago | |
.pre-commit-config.yaml | 10 months ago | |
CONTRIBUTING.md | 10 months ago | |
LICENSE | 1 year ago | |
MANIFEST.in | 11 months ago | |
README.md | 5 months ago | |
README_CN.md | 7 months ago | |
mkdocs.yml | 9 months ago | |
package.sh | 10 months ago | |
pyproject.toml | 10 months ago | |
requirements.txt | 7 months ago | |
results.png | 5 months ago | |
setup.py | 10 months ago |
LayoutXLM是一个用于多语言文档理解的多模态预训练模型,其旨在弥合视觉丰富文档理解的语言障碍。Vi-LayoutXLM在LayoutXLM的基础上上移除了基于ResNet x101 64x4d的视觉骨干网络,在不降低模型性能的同时提高了模型的训练和推理速度。
XFUND是一个多语言表单理解基准数据集,其中包括7种语言(中文、日语、西班牙语、法语、意大利语、德语、葡萄牙语)的带有键值对的人工标记表单。Vi-LayoutXLM使用其中文部分作为训练和测试集。数据集按如下格式存放:
datasets/XFUND
├── class_list_xfun.txt
├── zh_train
│ ├── image
│ │ ├── zh_train_0.jpg
│ │ ...
│ │ └── zh_train_99.jpg
│ └── train.json
└── zh_val
├── image
│ ├── zh_val_0.jpg
│ ...
│ └── zh_val_49.jpg
└── val.json
https://openi.pcl.ac.cn/wuzhf9/vilayoutxlm
代码目录结构遵循MindOCR官方仓库中的目录结构。
Ascend910 + MindSpore2.0.0 + Python3.8.0
详细训练超参数请查看./configs/kie/vi_layoutxlm/ser_vi_layoutxlm_xfund_zh.yaml
mpirun --allow-run-as-root -n 8 python tools/train.py --config configs/kie/vi_layoutxlm_xfund_zh.yaml
python tools/eval.py --config configs/kie/vi_layoutxlm_xfund_zh.yaml
https://arxiv.org/pdf/2104.08836
https://arxiv.org/pdf/2210.05391
https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.6/doc/doc_ch/kie.md
https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.6/doc/doc_ch/algorithm_kie_vi_layoutxlm.md
No Description
Python Markdown Text C++ Shell other
Dear OpenI User
Thank you for your continuous support to the Openl Qizhi Community AI Collaboration Platform. In order to protect your usage rights and ensure network security, we updated the Openl Qizhi Community AI Collaboration Platform Usage Agreement in January 2024. The updated agreement specifies that users are prohibited from using intranet penetration tools. After you click "Agree and continue", you can continue to use our services. Thank you for your cooperation and understanding.
For more agreement content, please refer to the《Openl Qizhi Community AI Collaboration Platform Usage Agreement》