Camera self-calibration network based on face shape estimation

Camera calibration is of great significance. However, camera calibration still faces significant challenges. Target-based camera calibration depends on the accuracy of the calibration board and the extraction of corner points. Traditional self-calibration methods rely on the matching accuracy of 2D points. Face shape estimation has made great progress. This paper proposes a new camera self-calibration framework based on face shape estimation. The neural network is used to predict the 3D shape of the face and camera parameters. BPnP algorithm is used to solve the PnP mapping to estimate the camera's external parameters. This method achieves the best performance in face 3D estimation and camera parameter estimation compared with related methods.