Sequential sketch identification method fusing texture features and shape features

A shape feature and texture feature technology, which is applied in the field of sequential sketch recognition and image classification tasks, can solve the problems of touching or ignoring the timing of sketches, and the lack of rich texture features, so as to achieve accurate training data, rich description ability, The effect of good accuracy

Active Publication Date: 2018-06-29
DALIAN UNIV OF TECH
View PDF7 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, it is a huge challenge to automatically recognize hand-drawn sketches with variable stroke order and different styles, so that the automatic recognition rate has reached the ceiling in recent years
Most of the current methods, especially those based on deep networks, do not ignore the geometric features of sketches, and use the texture features that have achieved great success i...

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
  • Sequential sketch identification method fusing texture features and shape features
  • Sequential sketch identification method fusing texture features and shape features
  • Sequential sketch identification method fusing texture features and shape features

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0038] For step 1, a sequence of images is acquired. Taking P=5 can effectively utilize the timing characteristics of sketch strokes, such as figure 2 shown.

[0039] By cropping and flipping horizontally, the image sequence (I 1 , I 2 ,...,I P ) expands to 10 images. image 3 , the 10 images on the left end show this effect: from right to left, the images on the odd bits are cropped on the original image, and the images on the even bits are cropped on the horizontally flipped image of the original image. The sequence is to keep the upper left, lower left, upper right, lower right and center part of the original image; 2 As an example, the 10 images are named from right to left as At this point, each original sketch S becomes P*10 images, among which from Image I t , k is an integer between [1,10].

[0040] image 3 , in order from right to left, 10 images are input into the GRU of the first stage for learning.

[0041] Remove the Coded Shape Context used to gene...

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 belongs to the field of computer vision and discloses a sequential sketch identification method fusing texture features and shape features. The method comprises the steps of firstly, according to a stroke sequence of a sketch, obtaining an image sequence; secondly, extracting the texture features and the shape features of images to form a feature sequence corresponding to the image sequence; and thirdly, inputting the features to a network containing two stages for performing learning, wherein in the first stage, two recurrent neural networks receive the texture features and theshape features of the images; and in the second stage, outputs in the previous stage are fused firstly, then a fused output is input to a third recurrent neural network, and finally a result is obtained through a classifier, so that iterative learning is performed according to the sequence in the sequences. The method has the advantages that a geometric descriptor is used for sketch identification, and the recurrent neural networks are adopted for performing effective learning on sequential features, so that the defect that an original identification model ignores the shape features and the sequential features of the sketch is remarkably improved and the sketch identification efficiency is better improved.

Description

technical field [0001] The invention belongs to the field of computer vision, relates to image classification tasks, in particular to a time-series sketch recognition method which combines texture and shape features. Background technique [0002] Sketches that reflect the main features of an object are an effective way for people to communicate ideas. However, it is a huge challenge to automatically recognize hand-drawn sketches with variable stroke order and different styles, so that the automatic recognition rate has reached the ceiling in recent years. Most of the current methods, especially those based on deep networks, do not ignore the geometric features of sketches, and use the texture features that have achieved great success in natural image recognition for sketch recognition, while ignoring that texture features are not very important in sketches. Abundant defects; it is to treat sketches as handwritten letters with a fixed structural order, ignoring the timing of...

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
IPC IPC(8): G06K9/46G06K9/62G06T5/50G06T11/00G06N3/04
CPCG06T5/50G06T11/001G06T2207/20221G06T2207/10016G06V10/44G06N3/045G06F18/23213
Inventor 贾棋樊鑫秦启炜唐国磊刘日升徐秀娟赵晓薇许真珍
Owner DALIAN UNIV 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