Method and apparatus for training a neural network, image recognition method and storage medium
A method and an apparatus for training a neural network, an image recognition method and a computer readable storage medium are disclosed. The neural network includes a first model and a second model. The method for training a neural network includes: acquiring a second image from a first image, wherein a quality of the second image is lower than that of the first image; inputting the first image into the first model of the neural network, and inputting the second image into the second model of the neural network; calculating an attention map and a gradient map of the first model and an attention map and a gradient map of the second model; constructing a loss function based on a matrix of a dot product of the gradient map and the attention map of the first model and a matrix of a dot product of the gradient map and the attention map of the second model; and training the neural network by minimizing the loss function.
1 . A method for training a neural network comprising a first model and a second model, the method comprising:
acquiring a second image from a first image, wherein a quality of the second image is lower than that of the first image;
inputting the first image into the first model of the neural network, and inputting the second image into the second model of the neural network;
calculating an attention map and a gradient map of the first model and an attention map and a gradient map of the second model;
constructing a loss function based on a square of a difference between a matrix of a dot product of the gradient map and the attention map of the first model and a matrix of a dot product of the gradient map and the attention map of the second model; and
training the neural network by minimizing the loss function.
2 . The method according to claim 1 , further comprising: after calculating of attention map, softening the attention map of the first model and the attention map of the second model.
3 . The method according to claim 2 , wherein the loss function is constructed as a square of a difference between a matrix of a dot product of the gradient map and a softened attention map of the first model and a matrix of a dot product of the gradient map and a softened attention map of the second model.
4 . The method according to claim 1 , wherein the first model and the second model are two symmetrical branches of the neural network.
5 . The method according to claim 4 , wherein the first model and the second model each comprise one or more convolutional layers and one or more fully connected layers.
6 . The method according to claim 1 , wherein the matrix is a Gram matrix.
7 . The method according to claim 1 , further comprising: training the neural network by using the loss function, a knowledge distillation loss function, and a classification loss function.
8 . The method according to claim 1 , wherein the first image and the second image include a face.
9 . An image recognition method, comprising:
inputting an image to be recognized into the second model of the neural network trained by the method according to claim 1 for recognition.
10 . An apparatus for training a neural network comprising a first model and a second model, comprising:
an acquisition means configured to acquire a second image from a first image, wherein a quality of the second image is lower than that of the first image;
an input means configured to input the first image into the first model of the neural network, and input the second image into the second model of the neural network;
a calculation means configured to calculate an attention map and a gradient map of the first model and an attention map and a gradient map of the second model; and
a construction means configured to construct a loss function based on a square of a difference between a matrix of a dot product of the gradient map and the attention map of the first model and a matrix of a dot product of the gradient map and the attention map of the second model,
wherein the neural network is trained by minimizing the loss function.
11 . The apparatus according to claim 10 , further including: a softening means configured to soften, after calculating of attention map, the attention map of the first model and the attention map of the second model.
12 . The apparatus according to claim 11 , wherein the loss function is constructed as a square of a difference between a matrix of a dot product of the gradient map and a softened attention map of the first model and a matrix of a dot product of the gradient map and a softened attention map of the second model.
13 . The apparatus according to claim 10 , wherein the first model and the second model are two symmetrical branches of the neural network.
14 . The apparatus according to claim 13 , wherein the first model and the second model each include one or more convolution layers and one or more fully connected layers.
15 . The apparatus according to claim 10 , wherein the matrix is a Gram matrix.
16 . The apparatus according to claim 10 , wherein the neural network is trained by using the loss function, a knowledge distillation loss function, and a classification loss function.
17 . The apparatus according to claim 10 , wherein the first image and the second image include a face.
18 . A non-transitory computer readable medium storing a program which can be executed by a processor to perform the method according to claim 1 .