A face recognition system that extracts feature vectors using a deep learning-based face recognition model according to an aspect of the present invention capable of extracting feature vectors that enable accurate face recognition without being dependent on place or lighting, input data A plurality of face image processing unit for processing the image to generate the output data; And a feature vector generating unit that merges output data output from the last face image processing unit among the plurality of face image processing units into a single layer to generate a predetermined number of feature vectors, and includes one of the plurality of face image processing units. The first face image processing unit inputs a face image as the input image, and the n+1th face image processing unit inputs the nth face image processing unit output data as the input image, and the plurality of face image processing units inputs the input data. A first unit to generate a feature map by applying a convolution filter; A second unit that weights the feature map generated by the first unit; And a calculator configured to add the feature map weighted by the second unit and the input data input to the first unit to generate the output data.
展开▼