the wise doc translator is a personal project where I aim to build a system that convert a document from a language to another keeping all the design (Layout, Logo, Sign, ... )
Video demo : Linkedin Post
OS : Ubuntu 18.04
CPU : Intel(R) Core(TM) i7-8750H CPU @ 2.20GHz
GPU : GeForce RTX 2070 with Max-Q Design
The current stable software version is : 1.0
pip install -r requirements.txt- PyTorch and its dependencies (torchvision, CUDA, ...) : https://pytorch.org/get-started/locally/
- install MMEditing and its dependencies (NCCL, mmcv, ...) : https://mmediting.readthedocs.io/en/latest/install.html Current working versions are :
- Python = 3.8.5
- PyTorch = 1.7.1
- torchvision = 0.8.2
- CUDA = 11.0
- NCCL = 2
git clone https://github.com/raoufkeskes/TheWiseDocTranslator.gitcd TheWiseDocTranslator- Copy your custom MMediting installation folder into the project :
cp -Rf SRC_PATH/mmediting/ ./ - Download pretrained :
- CRAFT
- STR (according to your config, for example we actually use TPS-ResNet-BiLSTM-Attn.pth)
- DeepFillv2 for Places2
-
Fix all your parameters on config.yaml such as : data_folder, results_folder, dest_lang, ...
-
run
python run_translation.py
[1] Y. Baek, B. Lee, D. Han, S. Yun, and H. Lee, “Character region awareness for text detec- tion,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 9365–9374, 2019.
[2] J. Baek, G. Kim, J. Lee, S. Park, D. Han, S. Yun, S. J. Oh, and H. Lee, “What is wrong with scene text recognition model comparisons? dataset and model analysis,” in International Conference on Computer Vision (ICCV), 2019.
[3] J. Yu, Z. Lin, J. Yang, X. Shen, X. Lu, and T. S. Huang, “Free-form image inpainting with gated convolution,” in Proceedings of the IEEE International Conference on Computer Vision, pp. 4471–4480, 2019.
[4] Z. Liu, P. Luo, X. Wang, and X. Tang, “Deep learning face attributes in the wild,” in Proceedings of International Conference on Computer Vision (ICCV), December 2015.
[5] B. Zhou, A. Lapedriza, A. Khosla, A. Oliva, and A. Torralba, “Places: A 10 million im- age database for scene recognition,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017.