Fast face angle recognition method based on deep learning
A deep learning and recognition method technology, applied in the field of face recognition, can solve the problems of uncertain face angle and low robustness, achieve fast and accurate face angle recognition, and improve the effect of face recognition accuracy
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0016] A fast face angle recognition method based on deep learning, the method includes the following steps: (1) mark the face images when the rotation is 0°, plus or minus 15°, plus or minus 30° and plus or minus 45° to establish Database; (2) The angle label is extracted through a five-layer convolution module, and converted into a 2048-dimensional feature vector after two layers of full connection, which is used as the input of the softmax classification layer for classification to establish a convolutional neural network structure; ( 3) Based on the ubuntu 16.04 operating system, under GPU1080, the convolutional neural network is trained under the deep learning framework CAFFE, and the face angle recognition model is obtained.
Embodiment 2
[0018] According to the fast face angle recognition method based on deep learning described in embodiment 1, the specific process of establishing the database is: the database is 300 different people's face images, a total of 7000, the images in the database Including face images when the head rotates 0°, plus or minus 15°, plus or minus 30° and plus or minus 45° around each axis in the three-dimensional vertical coordinate system, and the rotation angle is normalized to a value between 0-1 The value is used as the label value to mark the angle of the face in each image, and the marked image is used for the training of the neural network.
Embodiment 3
[0020] According to the fast face angle recognition method based on deep learning described in embodiment 1 or 2, the specific process of establishing the convolutional neural network structure is as follows: the input layer accepts input data, obtains image data and corresponding label values thereof, and establishes The data set contains three label values, which correspond to the angle labels of each axis rotation of the three-dimensional coordinate system, and then undergoes feature extraction through five layers of convolution modules. Each convolution module includes a convolution layer and a pooling layer, which will be extracted to The feature vector of the input to the full connection layer, after two layers of full connection, the feature map is converted into a 2048-dimensional feature vector, which is used as the input of the softmax classification layer for classification. The three labels correspond to three parallel classification layers. Each classification Th...
PUM
Abstract
Description
Claims
Application Information
- R&D Engineer
- R&D Manager
- IP Professional
- Industry Leading Data Capabilities
- Powerful AI technology
- Patent DNA Extraction
Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.
© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com