The HELEN* dataset is a rectified version of the original HELEN dataset.
You could download the dataset using ( Google Drive | Baidu Drive | Dropbox )
File structure:
train/ # the training set
xxx_image.jpg # the face image
xxx_label.png # the mask annotations
xxx_viz.jpg # annotation visualization
test/ # the testing set (in original size)
...
test_resize/ # the testing set (resized)
...
All images in train/
come from the original HELEN. We relabeled the hair and facial skin regions in train/
. Annotations in test/
also come from the original HELEN. While annotations in test_resize/
comply with the resized HELEN.
Mask annotation can be loaded by
import cv2
labels = cv2.imread('xxx_label.png', cv2.IMREAD_GRAYSCALE)
Meaning of annotation values:
0: background
1: facial skin
2: left brow (viewer side)
3: right brow
4: left eye
5: right eye
6: nose
7: upper lip
8: inner mouth
9: lower lip
10: hair
F scores can be calculated using f1_score.py
.
If you find the dataset useful, please consider citing
@inproceedings{lin2019face,
title={Face Parsing with RoI Tanh-Warping},
author={Lin, Jinpeng and Yang, Hao and Chen, Dong and Zeng, Ming and Wen, Fang and Yuan, Lu},
booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
pages={5654--5663},
year={2019}
}