Neural Network Object Detection Merging Position and Attribute Channels
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current object detection methods require significant computing resources and processing time, making them inefficient for real-time performance, especially on embedded terminals, as they separate position and attribute detection into distinct steps.
Innovation Solution
An integrated object detection method using a neural network that simultaneously detects position and attribute information, reducing computational requirements and enabling lightweight terminal deployment by sharing feature maps for both tasks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If position detection and attribute detection are separated into distinct steps, then detection accuracy can be maintained, but computing resources and processing time increase significantly
Solution Approach 1:
The patent merges position detection and attribute detection into a single integrated neural network model. The network simultaneously outputs both position information (bounding boxes) and attribute information (object characteristics) from shared feature maps, eliminating the need for separate detection steps while maintaining accuracy through unified multi-task learning
Solution Approach 2:
The neural network is designed with multi-functionality to perform both position detection and attribute detection tasks simultaneously. The same network architecture and feature extraction processes serve dual purposes, making the system universally capable of handling multiple detection objectives without requiring separate specialized models
2Reliability
If separate detection steps are used for position and attributes, then comprehensive detection can be achieved, but device complexity and computational requirements increase
Solution Approach 1:
The patent combines multiple detection functions into a single neural network architecture that processes images once and generates both position and attribute results simultaneously, reducing device complexity by eliminating redundant computational components and streamlining the detection pipeline
Solution Approach 2:
The patent discards the traditional sequential approach of separate detection steps and recovers computational efficiency by using shared feature maps that serve both detection tasks, thereby reducing overall computational requirements while maintaining comprehensive detection capabilities
3Measurement precision
If traditional object detection methods are used, then accurate results can be obtained, but real-time performance is not achieved on embedded terminals
Solution Approach 1:
The patent merges position and attribute detection into a single unified process that operates in real-time on embedded terminals by sharing computational resources and processing steps, thereby reducing processing time while maintaining accurate detection results through efficient multi-task learning
Data Source
AI summary
The present disclosure provides an object detection method, an object detection device, an object detection system and a storage medium. The object detection method includes: acquiring an image to be processed; and inputting the image to be processed into a neural network to obtain a feature map outputted by the neural network. The feature map includes position channels and attribute channels; the position channels include at least one group of candidate position information respectively corresponding to at least one candidate position of at least one prediction object in the image to be processed; and the attribute channels include at least one group of candidate attribute information respectively corresponding to the at least one candidate position.