Single-image face three-dimensional reconstruction method based on self-alignment double regression

• Main Authors: WU GANGSHAN, RUAN ZEYU, WANG LIMIN
• Format: Patent
• Language: Chinese; English
• Published: 30.08.2022
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS

The invention discloses a single-image face three-dimensional reconstruction method based on self-alignment double regression. The method comprises a training data processing stage, a network configuration stage, a training stage and a testing stage. According to the method provided by the invention, the unshielded area of the human face is enhanced based on the attention mechanism, so that the robustness of shielding is improved; according to the method, the posture and the shape of the three-dimensional face are decoupled, three-dimensional face shape regression irrelevant to the posture is independently learned, the face reconstruction precision is improved, and the negative influence of posture change is weakened; according to the method, the attitude information is encoded in a rough face model serving as an intermediate result, and the face attitude is extracted by combining the self-alignment post-processing process with the estimated visibility of the face region, so that compared with a direct attitu