Abstract: The present invention discloses an image processing method. The image processing method includes the following steps: (a), a to-be-processed image is corrected as a first correction image according to a first mapping relationship along a correction direction; (b) the first correction image by an angle is rotated; and (c) the rotated first correction image is corrected as a second correction image according to a second mapping relationship along the same correction direction. In embodiment, given that the to-be-processed image is deformed along two different directions, the to-be-processed image is corrected along the same correction direction, such that correction complexity could be reduced.

Abstract: A mixed-precision quantization method for a neural network is provided. The neural network has a first precision and includes several layers and an original final output. For a particular layer, quantization of second precision on the particular layer and an input is performed. An output of the particular layer is obtained according to the particular layer of second precision and the input. De-quantization on the output of the particular layer is performed, and the de-quantized output is inputted to a next layer to obtain a final output. A value of an objective function is obtained according to the final output and the original final output. Above steps are repeated until the value of the objective function of each layer is obtained. A precision of quantization for each layer is decided according to the value of the objective function. The precision of quantization is one of first to fourth precision.

Abstract: A facial image screening method is provided in the present disclosure. The facial image screening method includes the following steps: tracking a plurality of feature points of at least one facial image; analyzing each feature point to acquire a region information corresponding to each feature point, wherein the region information comprises an image quality information, a face pose information, and a blocking degree information; scoring the image quality information to obtain a first arbitration score; scoring the face pose information to obtain a second arbitration score; scoring the blocking degree information to obtain a third arbitration score; generating a comprehensive quality score according to the first arbitration score, the second arbitration score, and the third arbitration; and taking the plurality of feature points of the facial image as targets to be compared with captured features of a plurality of specific persons when the comprehensive quality score exceeds a threshold value.