Mixed cartoon image detection method based on deep learning
The invention discloses a mixed cartoon image detection method based on deep learning, and the method comprises the steps: firstly preparing a picture sample, preparing various types of images, and defining five classification tag cards to respectively correspond to different numerical values; divid...
Saved in:
Main Authors | , , |
---|---|
Format | Patent |
Language | Chinese English |
Published |
24.08.2021
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | The invention discloses a mixed cartoon image detection method based on deep learning, and the method comprises the steps: firstly preparing a picture sample, preparing various types of images, and defining five classification tag cards to respectively correspond to different numerical values; dividing the samples into a training set and a test set according to the proportion, then zooming pictures to the size of AxB, and converting the pictures into a matrix of AxBx3; converting corresponding labels into one-hot coding forms, converting a matrix into a multi-dimensional vector through a feature extraction layer, and then connecting the multi-dimensional vector with a full connection layer of a 256 hidden layer and a Dropout layer; connecting a full connection layer of five output neurons, using softmax as an activation function, using cross entropy as a target loss function, carrying out loss optimization through an adam optimizer, and then obtaining an optimal model; and inputting a to-be-detected image int |
---|---|
Bibliography: | Application Number: CN202110601966 |