Unlock instant, AI-driven research and patent intelligence for your innovation.

Deep learning detection method for salient region

A technology of deep learning and detection methods, applied in the fields of computer vision and image processing, can solve problems such as saliency map fusion, and achieve the effect of accurate saliency map and simple network structure

Pending Publication Date: 2020-09-04
BEIJING UNION UNIVERSITY
View PDF1 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Although this method considers multi-scale factors in the extraction of salient regions, this method finally selects a saliency map at one scale as the final saliency map, and does not actually fuse saliency maps at multiple scales

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
  • Deep learning detection method for salient region
  • Deep learning detection method for salient region
  • Deep learning detection method for salient region

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0051] Such as figure 1As shown, step 100 is executed to construct a multi-scale deep network. Including, performing step 101, bottom-up feature extraction. An image is input into the network, and the feature extraction module performs feature extraction from bottom to top to obtain multi-scale features. Using a 16-layer VGG network, the feature extraction module includes 13 convolutional layers, and performs nonlinear mapping and maximum pooling operations through the ReLU linear correction unit. The features of the four scales are expressed as {F 2 ;F 3 ;F 4 ;F 5}. Execute step 102, top-down feature connection. The features of different levels are connected, expressed as follows: Among them, d k Represents a deconvolution operation with a convolution kernel size of 4x4 and a step size of 2, f k Represents a 1x1 convolution operation, represents the feature map, F k-1 Represents the features of the k-1th layer of the VGG deep network structure, F k Represents th...

Embodiment 2

[0062] Multi-scale salient features have been successfully used in traditional salient region detection based on artificially designed features. However, in CNN-based methods, multiple deep networks need to be trained for multi-scale feature extraction. Although good performance has been achieved, the calculation and storage costs are huge, and it is not practical in practical applications. However, recent work has shown that deep convolutional neural networks are inherently multi-scale feature hierarchies. Therefore, the feature maps of the sub-sampled convolutional layers in different spaces can be regarded as multi-scale features, and do not need to be constructed through additional network structures. Influenced by this idea, the present invention proposes a multi-scale deep network for salient region detection, which utilizes the inherent hierarchical features of deep networks. The present invention inputs RGB images into a deep convolutional neural network, and predict...

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 provides a deep learning detection method for a salient region, and the method comprises the steps: constructing a multi-scale deep network, and carrying out the training of the multi-scale deep network; and performing saliency prediction and fusion. The invention provides a deep learning detection method for a salient region. A multi-scale deep network is used for salient region detection, hierarchical features in the deep network are utilized, an RGB image is input into a deep convolutional neural network, and a feature map under different scale spaces is predicted through bottom-to-top feature extraction and top-to-bottom feature integration.

Description

technical field [0001] The invention relates to the technical field of computer vision and image processing, in particular to a deep learning detection method of a salient area. Background technique [0002] The purpose of saliency detection is to find salient regions in images, which is a fundamental problem in image understanding and analysis, and is often used as preprocessing for other computer vision problems, such as image classification, image segmentation, and relocalization. Although saliency detection has made great progress in recent years, saliency is still a challenging task. [0003] Traditional saliency detection methods usually use multiple saliency cues or prior information, such as local or global contrast, boundary priors. These methods cannot identify and understand semantic object concepts in images due to the use of low-level artificially designed features and models. Recently, deep convolutional neural networks have achieved impressive results in vis...

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): G06K9/32G06N3/04
CPCG06V10/25G06N3/045
Inventor 梁晔马楠李文法张磊徐俊李大伟孙晨昊周航王楠
Owner BEIJING UNION UNIVERSITY