Lightweight face living body detection method and system based on dual auxiliary supervision

The invention discloses a lightweight human face in-vivo detection method and system based on dual auxiliary supervision. The detection method comprises the following steps: S1, preprocessing a target image to form a preprocessed image; s2, extracting features of different scales according to the preprocessed image, and fusing the features of different scales through an attention mechanism to obtain universal semantic features; s3, reconstructing depth features and reflection features of the face in the real scene according to the universal semantic features; s4, fusing the general semantic feature, the depth feature and the reflection feature to obtain a discrimination feature; and S5, outputting a judgment result according to the judgment features. By extracting the features of different sizes and dimensions, face fraud attack lines in the features of different sizes and dimensions can be mined; features of different sizes and dimensions are fused through an attention mechanism, the receptive field and deta