Human face detection method and human face detection device
A technology of face detection and target detection, applied in the field of face recognition, which can solve the problems of relying on face detection for pose estimation, high error rate of face detection, low accuracy of face features, etc., achieve rich description and improve task processing High efficiency and high accuracy
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
no. 1 example
[0041] refer to figure 1 , which shows a flow chart of the steps of an embodiment of a face detection method of the present invention, which may specifically include the following steps:
[0042] Step 101, using the pre-trained convolutional neural network model to extract multiple face features of different levels of networks from the face image to be tested, and obtain multiple face feature vectors corresponding to different levels of networks;
[0043] Among them, the convolutional neural network model has multiple different-level networks, and the information contained in the features extracted by different-level networks is also hierarchically distributed. The features extracted by the low-level network focus on describing edges and corners, including better positioning features. , so it is suitable for learning tasks such as pose estimation; the features extracted by the high-level network are features related to the corresponding category, so it is suitable for learning...
no. 2 example
[0054] On the basis of the above embodiments, this embodiment further discusses the face detection method of the present invention.
[0055] Before using the above-mentioned convolutional neural network model to perform feature extraction on the face, the convolutional neural network model needs to be trained, and before the training, the method according to the embodiment of the present invention also needs to prepare training samples.
[0056] The embodiment of the present invention utilizes the training set of the wide face (WIDER FACE) database to generate training samples for face detection and pose estimation. WIDER FACE contains a total of 3,2203 images and 39,3703 face annotations. Divided into 61 scenes, such as parades, parties, festivals, conferences, etc. In each scenario, 40%, 10%, and 50% of the samples can be randomly selected as training samples, verification samples, and test samples, respectively. Moreover, in the above annotations, in addition to the face ...
no. 3 example
[0127] On the basis of the above-mentioned embodiment, combine below Figure 4 Continue to discuss the face detection method of the embodiment of the present invention.
[0128] In order to detect faces of different sizes, in one embodiment, before step 101 is performed, a face pyramid can also be generated from the face image to be tested, which specifically includes: using an image pyramid method to perform scaling processing on the face image to be tested to obtain A plurality of human face images to be tested of different sizes belonging to the same original image; the plurality of human face images to be tested are sequentially input into a pre-trained convolutional neural network model for face detection.
[0129] In this example, a Figure 5A The image to be tested shown is enlarged to 6 times. Since the size of the training sample is 224×224, a human face with a minimum size of 37×37 can be detected at this time, and then the enlarged image is gradually reduced until ...
PUM
Login to view more Abstract
Description
Claims
Application Information
Login to view more - 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.
© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap