BP4D stands for BP4D-Spontaneous dataset, which is a 3D video dataset containing natural facial expressions of different young people.

This dataset uses well-validated emotion induction to elicit emotional expressions and paralinguistic communication. Frame-level ground truth of facial actions is obtained using a facial action coding system. Facial features are tracked in 2D and 3D domains using person-specific and generic methods.

The dataset includes 41 participants (23 females, 18 males) with an age range of 18 to 29 years old; 11 of them are Asian, 6 are African American, 4 are Hispanic, and 20 are European American. The emotion elicitation protocol is designed to effectively elicit emotions in participants. The eight tasks consist of an interview process and a series of activities that elicit eight emotions.

The database is built by participants. Each participant is associated with 8 tasks. Each task has 3D and 2D videos. In addition, metadata includes manually annotated action units (FACS AU), automatically tracked head poses, and 2D/3D facial landmarks.