Neural network training method and device, equipment and storage medium
A neural network training and network technology, applied in the field of deep learning, can solve the problem of low network accuracy for students
Active Publication Date: 2019-08-23
MEGVII BEIJINGTECH CO LTD
View PDF5 Cites 19 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0005] Based on this, it is necessary to provide a neural network training method, device, equipment and storage medium for the problem of low accuracy of the student network
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0048] The neural network training method, device, equipment and storage medium provided by this application are aimed at solving the problem of low accuracy of the student network. The technical solution of the present application and how the technical solution of the present application solves the above technical problems will be described in detail below through embodiments and in conjunction with the accompanying drawings. The following specific embodiments may be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.
[0049] The neural network training method provided in this embodiment can be applicable to such asfigure 1 shown in the application environment. The neural network training method described above is applied to the neural network training terminal. Neural network training terminals can be, but are not limited to, various personal computers, laptops, smart phones, tablets, and portable wearable device...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention relates to a neural network training method, device and equipment and a storage medium. By means of a preset extraction method, a terminal obtains first foreground information and secondforeground information of the to-be-trained picture, and according to the first foreground information, a teacher network low-dimensional feature map is obtained by adopting a preset encoder; meanwhile, according to the second foreground information, a student network low-dimensional feature map is obtained by using an encoder; and a supervised loss function of the student network is determined according to the teacher network low-dimensional feature map and the student network low-dimensional feature map, and training the student network according to the supervised loss function and a presetself-learning loss function, the self-learning loss function being used for training the student network according to the real data label. Therefore, a supervised loss function only acting on the foreground information and a self-learning loss function obtained according to the real data label are used in the loss function of the training student network at the same time, so that the loss function of the student network is more accurate, and the precision of the student network is improved.
Description
technical field [0001] The present invention relates to the technical field of deep learning, in particular to a neural network training method, device, equipment and storage medium. Background technique [0002] Convolutional neural networks are usually used for data processing in computer vision tasks. Generally speaking, a network with higher accuracy requires more calculations, and is not suitable for small devices or mobile devices, etc. With the help of a large network, a small network is trained so that the small network also has relatively high accuracy in vision tasks. In this case, the large network is called the teacher network, and the small network is called the student network. During the training process, the student network not only obtains information from the real labels, but also obtains information from the teacher network. The accuracy of the student network trained in this way is often higher than that of the student network trained only with real lab...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Applications(China)
IPC IPC(8): G06N3/04G06N3/08G06K9/62
CPCG06N3/084G06N3/045G06F18/24
Inventor 郭义袁野王剑锋俞刚
Owner MEGVII BEIJINGTECH CO LTD