ototadana/occlusion-aware-face-segmentation

Occlusion-aware face segmentation

A model for occlusion-aware face segmentation.

This model was created following the procedures in mmsegmentation's PR [Feature] Support Delving into High-Quality Synthetic Face Occlusion Segmentation Datasets #2194.

For more information, see:

• https://github.com/open-mmlab/mmsegmentation/pull/2194/files
• https://github.com/kennyvoo/face-occlusion-generation

How to use

Use with mmsegmentation.

Example:

from mmseg.apis import inference_model, init_model, show_result_pyplot
import mmcv

config_file = 'deeplabv3plus_r101_512x512_face-occlusion.py'
checkpoint_file = 'deeplabv3plus_r101_512x512_face-occlusion-93ec6695.pth'
model = init_model(config_file, checkpoint_file, device='cuda:0')

img = 'face-image.png'
result = inference_model(model, img)
show_result_pyplot(model, img, result, show=True, out_file='result.jpg', opacity=0.5)