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

Shadow detection method based on bidirectional multilevel feature pyramid

A feature pyramid and shadow detection technology, which is applied in neural learning methods, instruments, biological neural network models, etc., can solve problems such as insufficient training data, and achieve the effect of improving accuracy and improving accuracy

Pending Publication Date: 2022-08-05
GUANGZHOU UNIVERSITY
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] Compared with the traditional shadow detection method, the above deep learning shadow detection method has significantly improved accuracy, but there are still deficiencies. In the face of insufficient training data, it is necessary to fully mine the higher-level information of the data to get better results. the accuracy of

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
  • Shadow detection method based on bidirectional multilevel feature pyramid
  • Shadow detection method based on bidirectional multilevel feature pyramid
  • Shadow detection method based on bidirectional multilevel feature pyramid

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0036] A shadow detection method based on a bidirectional multi-level feature pyramid, comprising the following steps:

[0037] S1. Build a residual refinement module

[0038] progressively learn the features of each layer by taking the feature maps of two adjacent layers as input;

[0039] S2. Build a weighted feature fusion module

[0040] Make full use of features at different scales by learning weight parameters;

[0041] S3. Build a bidirectional multi-level feature pyramid network

[0042] The information of different layers is processed through two paths in different directions: one is from deep to shallow, and the resulting fusion layer is used for shallow to deep processing, and the other is in the opposite direction, and weighted feature fusion is used to fuse information in both directions;

[0043] S4. Perform model training

[0044] Adjust the model parameters according to the loss function.

[0045] In step S1, the feature maps of two adjacent layers are ta...

Embodiment 2

[0054] A shadow detection method based on a bidirectional multi-level feature pyramid, comprising the following steps:

[0055] S1. Build a residual refinement module

[0056] progressively learn the features of each layer by taking the feature maps of two adjacent layers as input;

[0057] S2. Build a weighted feature fusion module

[0058] Make full use of features at different scales by learning weight parameters;

[0059] S3. Build a bidirectional multi-level feature pyramid network

[0060] The information of different layers is processed through two paths in different directions: one is from deep to shallow, and the resulting fusion layer is used for shallow to deep processing, and the other is in the opposite direction, and weighted feature fusion is used to fuse information in both directions;

[0061] S4. Perform model training

[0062] Adjust the model parameters according to the loss function.

[0063] In step S1, the feature maps of two adjacent layers are ta...

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 relates to the technical field of shadow detection, and discloses a shadow detection method based on a bidirectional multi-level feature pyramid, which comprises the following steps: constructing a residual error refining module, gradually learning features of each layer by taking feature maps of two adjacent layers as input, constructing a weighted feature fusion module, and obtaining a weighted feature fusion module; the features of different scales are fully utilized by learning weight parameters, a bidirectional multi-level feature pyramid network is constructed, and model training is carried out. According to the shadow detection method based on the bidirectional multilevel feature pyramid, a residual error refining module is constructed; constructing a weighted feature fusion module; constructing a bidirectional multi-stage feature pyramid network, adding a residual error refining module into the feature pyramid network, and performing multi-stage feature fusion on the input feature map from two directions; according to the method, semantic information of high-level features and fine-grained features of low-level features are utilized, the accuracy of model prediction is improved through multi-level detection, and the method can be widely applied to the technical field of shadow detection.

Description

technical field [0001] The invention relates to the technical field of shadow detection, in particular to a shadow detection method based on a bidirectional multi-level feature pyramid. Background technique [0002] Shadow detection has always been a basic but difficult challenge in computer vision. Shadow detection provides the possibility to obtain the shape and position of objects in the image and restore the unlit environment. At the same time, the existence of shadows also brings obstacles to further understanding of images, so shadow detection plays a crucial role in object detection and tracking, semantic segmentation, intelligent driving and other fields. [0003] The existing shadow detection techniques mainly fall into two categories: one is the traditional shadow detection method, and the other is the shadow detection method based on deep learning. Traditional shadow detection methods mainly use color, gradient, texture and other features for shadow detection. Fo...

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): G06V10/80G06V10/82G06N3/04G06N3/08
CPCG06V10/806G06V10/82G06N3/08G06N3/048G06N3/045
Inventor 曹忠陈俊全尚文利赵文静王锋邓辉梅盈
Owner GUANGZHOU UNIVERSITY