search

YexingWan / Fast-Portrait-Segmentation

The MNN base implementation of SINet for CPU realtime portrait segmentation
MIT License
63 stars 13 forks source link

Sinet training code and dataset #3

Closed ilwoonam75 closed 3 years ago

ilwoonam75 commented 3 years ago

Thank you for your repository. Can you share your training code and dataset? If not, can you explain how to train sinet?

YexingWan commented 3 years ago

Sorry for slow reply. Here is the origin SINet implementation https://github.com/clovaai/ext_portrait_segmentation
Further, the paper link is also provided in readme. The weight file provided is the original trained weight in origin SINet implementation after conversion (.pth -> .onnx -> .mnn) The training dataset is EG1800 and an augmentation dataset generated by Baidu's DeepFashion dataset. The way to do dataAug can be found in paper as well. You can learn the way to train SINet by code and paper as well. Thank you for your question