Description:
We use the SmartToF camera to capture a novel RGB-D database SmartToFFace under a wide variety of complex scenarios from a total of 42 volunteers.
The images of each person in the database are organized in 5 video sequences corresponding to neutral frontal, pose variation, expression change, illumination and partial occlusion. In addition, 10 subjects are selected to capture an additional backlight shadow video sequence. Each capturing sequence contains 150 synchronized RGB-D image pairs, resulting in 33000 samples in total.
With the current epidemic of Coronavirus, it is important to perform accurate face recognition under the scenario of wearing masks. Thus, we add a large number of samples of volunteers wearing masks in different postures in the occlusion subset. In addition, we also add more challenging scenarios to our database, such as backlight shadows, extreme illumination, etc.
Application:
If you are interested in getting the database and can prove that your use is for academic purpose only, please send your “name + position + organization” to “zhenghaoyuan980108@sjtu.edu.cn”. Then we will review your information, and after the review is approved, we will send you a usage agreement by email.
