Face reconstruction method and system based on single picture

The invention discloses a face reconstruction method and system based on a single picture, and relates to the technical field of intelligent face reconstruction. The method comprises: acquiring a to-be-reconstructed face image; performing image feature extraction on the to-be-reconstructed face image to obtain a fused semantic face shallow feature map; and based on the fused semantic face shallow feature map, obtaining a generated face 3D model. Therefore, the three-dimensional face model can be learned and inferred from a single picture without depending on a pre-constructed shape model. 一种基于单张图片人脸重建方法及系统，涉及智能化人脸重建技术领域；其获取待重建人脸图像；对所述待重建人脸图像进行图像特征提取以得到融合语义人脸浅层特征图；以及，基于所述融合语义人脸浅层特征图，得到生成人脸3D模型。这样，可以从单张图片中学习和推理出三维人脸模型，无需依赖于预先构建的形状模型。