The system leverages a retrieval-based approach to self-portrait transformation. During training, given a target neutral pose, it searches for a matching selfie in the aforementioned database.