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

A Recognition Method Based on Hand-drawn Sketch

A recognition method and sketch technology, applied in character and pattern recognition, still image data retrieval, metadata still image retrieval, etc., can solve the problem of inability to realize the semantic understanding of sketches, and achieve accurate and fast retrieval results, multiple database categories, and retrieval. Precise results

Active Publication Date: 2020-06-02
TIANJIN UNIV
View PDF2 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Content-based retrieval methods do not learn from sketches, so they usually cannot achieve semantic understanding of sketches

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 Recognition Method Based on Hand-drawn Sketch
  • A Recognition Method Based on Hand-drawn Sketch
  • A Recognition Method Based on Hand-drawn Sketch

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0033] A recognition method based on hand-drawn sketches, see figure 1 , the method includes the following steps:

[0034] 101: Adjust each sketch in the initial category set according to the preset size, uniformly extract several interest points, and extract a square block for each interest point;

[0035] 102: Extract pixel point gradients in each square block, and quantify to units in 4 directions according to the direction, as the local features of each square block; use k-means clustering method to construct a visual dictionary, and each sketch uses 500-dimensional vector representation;

[0036] 103: Using intra-class clustering analysis and performing dimensionality reduction processing on the feature vectors of each category to obtain the classified database;

[0037] 104: Match the query sketch with the classified database to obtain the final retrieval result.

[0038] Wherein, the initial category set in step 101 is specifically:

[0039] A total of 20,000 sketch...

Embodiment 2

[0046] The scheme in embodiment 1 is further introduced below in conjunction with specific calculation formulas and examples, see the following description for details:

[0047] 201: Obtain a total of 20,000 sketch datasets, make keywords cover most common categories through preprocessing, and use the preprocessed sketch dataset as the initial category set;

[0048] Wherein, the step 201 specifically includes:

[0049] 1) First construct a sketch data set of 20,000 pieces as the basis for learning, evaluation and application;

[0050] 2) Extract 1000 most common labels from LabelMe (label library), and manually remove duplicates and non-compliance rules with 1000 common labels as a benchmark (set according to the needs in practical applications, the embodiment of the present invention There is no limit to this, e.g. labels for chairs, mugs);

[0051] 3) Supplement the deleted draft dataset with keywords through preset standards and keywords in the preset dataset, and use the...

Embodiment 3

[0101] Below in conjunction with concrete experimental data, mathematical formula, the scheme in embodiment 1 and 2 is carried out feasibility verification, see the following description for details:

[0102] experimental report

[0103] 1. Database

[0104] A collection of 20,000 hand-drawn sketches databases that are the basis for learning, evaluation and application. Extract 1000 most common labels from LabelMe, manually remove duplicate and irregular labels, and use this as the initial category set. Increase the number of categories by keywords from PSB and Caltech 256 datasets. Finally, keywords are artificially supplemented, and finally 250 keywords are obtained, and these words cover most of the common categories.

[0105] Using the full 250 classes, each sketch was converted to a grayscale bitmap and sized to 256×256. 28×28=784 local features are extracted for each sketch, and one feature is extracted in each grid. In order to create a visual dictionary, a huge numb...

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 hand-drawn sketch-based recognition method. The recognition method comprises the following steps of: adjusting each sketch in an initial category set according to a preset size, uniformly extracting a plurality of interest points and extracting a square block of each interest point; extracting a pixel point gradient in each square block, quantifying units in four directions according to directions and taking the units as local features of each square block; constructing a visual dictionary by using a k-means cluster method and expressing each sketch by using a 500-dimensional vector; obtaining a classified database by adopting in-category cluster analysis and carrying out dimensionality reduction on a feature vector of each category; and matching a queried sketchwith the classified database to obtain a final retrieval result. According to the method, sketches are used as visual input manners and robust visual local features are designed for the sketches, so that the recognition correctness is improved.

Description

technical field [0001] The invention relates to the field of image retrieval, in particular to a recognition method based on hand-drawn sketches. Background technique [0002] content based retrieval [1] The difference is that the user input for sketch-based retrieval is a simple binary sketch. Content-based retrieval methods do not learn from sketches, so they usually cannot achieve semantic understanding of sketches. Retrieval results are based purely on geometric similarity between sketch and image content [2][3] . [0003] An image synthesis system based on sketch retrieval that allows users to create novel and realistic images. Synthetic systems based only on sketches must rely on large amounts of data to counteract the problem of geometric disparity between sketch and image content [4] , or require the user to add a text label to the sketch [5] . The proposed system using template matching to recognize faces can help users get the correct proportions when drawin...

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): G06F16/583G06F16/58G06K9/32G06K9/46G06K9/62
Inventor 聂为之邓宗慧苏育挺
Owner TIANJIN UNIV
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

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com