Welcome to the webpage of the FAce Semantic SEGmentation (FASSEG) dataset.
The FASSEG dataset is made available here in two versions:
V1 contains 70 labeled faces and the original RGB images. This is exactly the same dataset we used in our paper . Faces are mainly taken from the MIT-CBCL and FEI datasets. Images are organized in two folders - train and test - matching the division we adopted in the paper.
V2 is a "high-precision V1". It contains the same images as the V1 but it's much more precise in the segmentations (see images below). We also introduced a different convention for the "nose" class, which is extended up to the level of the eyes.
Our advice is: if you need to compare with our results, choose V1. If you need a dataset to build your models, choose V2.
In any case, if you use our datasets, please cite our work .
 Massimo Mauro, Khalil Khan, Riccardo Leonardi, "Multi-class semantic segmentation of faces", International Conference on Image Processing (ICIP), 2015
 MIT Center for Biological and Computational Learning (CBCL), MIT-CBCL database, http://cbcl.mit.edu/software-datasets/FaceData2.html
 Centro Universitario da FEI, FEI database, http://www.fei.edu.br/~cet/facedatabase.html