Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

A sketch image processing method and system based on knowledge distillation

A processing method, a technology of sketches, which is applied in the knowledge distillation-based sketch processing method and its system field, can solve problems such as low accuracy rate, eliminate modal differences, solve low accuracy rate, improve classification ability and generalization sexual effect

Active Publication Date: 2021-11-02
CHENGDU KOALA URAN TECH CO LTD
View PDF12 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] In view of this, the present invention provides a sketch image processing method and system based on knowledge distillation, which solves the problem of low accuracy in existing sketch image retrieval methods

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 more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • A sketch image processing method and system based on knowledge distillation
  • A sketch image processing method and system based on knowledge distillation
  • A sketch image processing method and system based on knowledge distillation

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0023] In order to enable those skilled in the art to better understand the technical solutions of the present invention, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0024] like figure 1 As shown, a sketch processing method based on knowledge distillation, including:

[0025] S1: Obtain a pre-trained teacher model composed of image classification models, and extract teacher features and teacher classification probabilities of sample images based on the teacher model.

[0026] Further, the method for obtaining a pre-trained teacher model composed of an image classification model may be: initialize the teacher model based on SE-ResNet-50; obtain the manually drawn sample sketch and the sample sketch collected by the imaging unit The corresponding sample images constitute a training data set; train the teacher model based on the training data set; expand the pre-trained SE-ResNet-50, th...

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

No PUM Login to View More

Abstract

The invention discloses a knowledge distillation-based sketch image processing method and its system, comprising: obtaining a pre-trained teacher model composed of an image classification model, and extracting teacher features and teacher classification probabilities of sample images based on the teacher model; A student model for image classification based on the sketch image, and extract the first student feature and the first student classification probability of the sample image based on the student model, and extract the second student feature and the second student classification of the sample image corresponding to the sample sketch image based on the student model Probability; Computes classification loss and distillation loss based on teacher features, teacher classification probabilities, first student features, first student classification probabilities, second student features, and second student classification probabilities, and updates the student model based on classification losses and distillation losses; student The model generates zero-sample cross-modal retrieval results based on the sketch to be detected and the image to be queried.

Description

technical field [0001] The invention relates to the technical field of machine learning, in particular to a knowledge distillation-based sketch image processing method and system thereof. Background technique [0002] With the ubiquity of touchscreen devices, freehand sketches are becoming more and more accessible. Although the sketch graph exhibits a high degree of abstraction, it still contains enough structural details to describe its corresponding categories. Therefore, image retrieval based on sketch image has become a research hotspot in the field of machine learning technology. [0003] However, the existing sketch-based image retrieval methods all presuppose that the categories contained in the sketch to be queried and the image to be queried are exactly the same as those contained in the training model, without considering the sketch to be queried and The case where the queried image contains a category that was not included in the categories when the model was tr...

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
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G06F16/583G06F16/55
CPCG06F16/583G06F16/55
Inventor 徐行田加林沈复民申恒涛
Owner CHENGDU KOALA URAN TECH CO LTD
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products