Zero sample sketch retrieval method based on discriminative improvement

A discriminative, sketchy technique, applied to still image data retrieval, computer parts, character and pattern recognition, etc., which can solve problems such as poor performance

Inactive Publication Date: 2021-06-25
JINLING INST OF TECH
View PDF0 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the SBIR method does not perform well in the real world

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
  • Zero sample sketch retrieval method based on discriminative improvement
  • Zero sample sketch retrieval method based on discriminative improvement
  • Zero sample sketch retrieval method based on discriminative improvement

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0041] Below in conjunction with accompanying drawing and specific embodiment the present invention is described in further detail:

[0042] The present invention proposes a zero-sample sketch retrieval method based on discriminative enhancement, and trains the model by using a teacher network and training data. Then, the trained model is used to extract the feature vectors of the sketch to be queried and the image to be retrieved, and by comparing the similarity of these feature vectors, the zero-sample sketch retrieval problem is solved.

[0043] Among them, the flow chart of zero-shot sketch retrieval based on discriminative promotion is as follows: figure 1 As shown, the difference between SBIR and ZS-SBIR is as follows figure 2 As shown, the model architecture of zero-shot sketch retrieval based on discriminative boosting is shown in image 3 shown.

[0044] In the following, the network structure of the present invention based on PyTorch is taken as an example, and t...

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 zero sample sketch retrieval method based on discriminative improvement, and provides a teacher-student network system structure which is composed of a teacher network using a pre-training model and a student network guided by the teacher network to output. The method comprises the steps: employing a stronger pre-training model as a teacher network, and increasing a hard coding distance based on a prediction probability to further enhance the discrimination of the teacher network; taking the output of the teacher network as a learning target, and the student network is finely adjusted. through training, obtaining a sketch network model with better performance; during retrieval, inputting a sketch to be queried and each candidate image into the sketch network model to obtain feature vectors of the sketch to be queried and the candidate images; calculating the Euclidean distance between the feature vectors and measuring the similarity between the sketch query and each candidate image; and according to the similarity, enabling the model to return an image most similar to the sketch to be queried. The method has the advantages of high retrieval accuracy, high model stability and high applicability.

Description

technical field [0001] The invention relates to the field of sketch-based image retrieval, in particular to a zero-sample sketch retrieval method based on discriminative enhancement. Background technique [0002] Sketch-Based Image Retrieval (SBIR) technology is widely used in many practical applications, such as animation, e-commerce and security fields. It allows users to use freehand sketches instead of traditional text and images as input to search for images of interest. [0003] Given a sketch to be queried, the purpose of the SBIR task is to retrieve images in the target dataset that have similar semantics to the queried sketch. To this end, a training dataset of labeled sketches and images is needed so that the model can learn the semantic relationship between the sketch domain and the image domain. [0004] In general, the training dataset and the target dataset in the SBIR task share all sample categories. That is, the categories of the retrieved images are alre...

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 Applications(China)
IPC IPC(8): G06F16/53G06K9/62
CPCG06F16/53G06F18/22G06F18/24
Inventor 赵海峰吴天健张燕
Owner JINLING INST OF TECH
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

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