Rotation invariance face detection method and device, readable storage medium and equipment

Abstract

The invention discloses a rotation invariance face detection method and device, a readable storage medium and equipment, and belongs to the field of mode recognition. Comprising the following steps: a face image is input into a three-level convolutional neural network to obtain a face position and a face attribute; the first-level convolutional neural network classifies the face image to a first angle range or a second angle range according to the attitude angle, and rotates the face image in the second angle range to the first angle range; the second-level convolutional neural network classifies the face images in the first angle range into a third angle range, a fourth angle range or a fifth angle range according to attitude angles, and rotates the face images in the fourth angle range or the fifth angle range to the third angle range; and the third-level convolutional neural network processes the face image in the third angle range to obtain a face position and a face attribute. According to the invention, the problems of detection and attribute judgment of the rotating face are effectively solved, and the invention has relatively low time complexity and better face detection and attribute judgment effects.

Description

technical field [0001] The invention relates to the field of pattern recognition, in particular to a rotation invariant human face detection method, device, computer-readable storage medium and equipment. Background technique [0002] Convolutional Neural Networks (CNN) have made great progress in the field of pattern recognition, especially in the field of image face recognition, where the recognition effect can far exceed that of human eyes. Since CNN can adaptively obtain image features and obtain better classification results, CNN-based face detection has also made a major breakthrough. However, in order to obtain a higher detection rate and a lower error detection rate, the existing face detection methods often use complex and deep network structures, which leads to the inability of CNN-based face detection to achieve real-time detection. the goal of. [0003] The problems of low recall rate of face detection and inaccurate judgment of face attributes due to face rota...

AI Technical Summary

Problems solved by technology

[0005] This method requires multiple detection and attribute judgment operations on the same image, which has a high time complexity.

If you want to obtain accurate detection results, you need to reduce the rotation interval a of the image, which will lead to greater time complexity

In addition, due to the diversity of rotation angles of an image face, a needs to traverse a wider range of angles, which will result in greater time overhead

Images