*NEW* Mut1ny is making part of its head/face segmentation dataset available for free for non-commercial purposes under the Attribute-Non-Commercial-ShareAlike4.0 International license
This new arrangement of the dataset contains over 16.5k (16557) fully pixel-level labeled segmentation images. Facial images are included from different ethnicity, ages and genders making it a well balanced dataset. Also there is a wide variety of facial poses and different camera angles to provide a good coverage from -90 to 90 degrees facing. Some images even contain multiple head/face segmentation depending on if the second or third face takes up enough screen real estate space.
For each real image there exist a PNG RGB label image pair. It encodies the 11 different labeled areas of the face using the following RGB labels:
|RED VALUE||GREEN VALUE||BLUE VALUE||FACIAL LABEL|
An XML file for instant training setup is also included.
This dataset was compiled by Mut1ny and the help of nearly 100 other voluntary people who added labeled data via data submission. For their help in the past we are very grateful.
How to obtain?
You can download the free dataset by just submitting the form below. But please note by submitting the form you agree to be contacted by Mut1ny for potential questions and commercial purposes
If you are interested in commercial usage and also want to obtain a larger version of the dataset which has more than 2.5 times the amount of labeled data (over 41500 pixel labeled facial images as registered in October 2020). You can either become a commercial or academic or higher level Patreon. Alternatively, if you do not like to use Patreon or their policies and process you can simply contact us with the same form above. We will try to sort out an alternative arrangement that better suits you.
If you do not have the resources or time to setup, find and train the right architecture
Why not checkout our Faceprocessor suite it offers a ready to plug-in solution, fully trained tried and tested segmentation network and even more.