YOLO-V2 Neural Network Segmentation for Edge Device Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing YOLO-V2 neural network requires high computational resources and has relatively low detection accuracy, making it challenging for edge computing devices with limited computing power to achieve efficient object detection.
Innovation Solution
The improved YOLO-V2 neural network incorporates a residual block, decomposes convolution layers into smaller ones with fewer filters, and removes the reorganization layer, reducing computational burden while enhancing identification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large and deep neural network framework is adopted to achieve favorable detection effects, then detection accuracy is improved, but computational burden and memory storage capacity requirements are significantly increased
Solution Approach 1:
The patent divides the original YOLO-V2 convolution layer with many filters into multiple smaller convolution layers with fewer filters each. This segmentation reduces the computational burden of each layer while maintaining the overall detection accuracy through the residual block architecture that preserves important feature information across the divided layers.
2Measurement precision
If a large and deep neural network framework is adopted to achieve favorable detection effects, then detection accuracy is improved, but memory storage capacity requirements are significantly increased
Solution Approach 1:
The patent segments the deep neural network into multiple shallower convolution layers, each requiring less memory storage. By dividing the large convolution layer into smaller layers with fewer filters, the peak memory usage is reduced while the residual connections ensure that important features are preserved across the segmented structure.
3Reliability
If the original YOLO-V2 neural network is used, then it provides a baseline detection capability, but it is difficult to apply to edge computing devices with less computing power
Solution Approach 1:
The patent transforms the original YOLO-V2 architecture into a form suitable for edge devices by segmenting large convolution layers into smaller layers. This reduces the computational requirements and memory usage, making the network adaptable to edge computing devices with limited resources while maintaining detection capability through the residual block design.
Solution Approach 2:
The patent changes the parameters of the convolution layers by reducing the number of filters in each layer while maintaining the overall network depth through residual connections. This parameter adjustment allows the network to function effectively on edge devices with less computing power while preserving the essential detection capabilities of the original YOLO-V2.
Data Source
AI summary
An object detection device and an object detection method based on a neural network are provided. The object detection method includes: receiving an input image and identifying an object in the input image according to an improved YOLO-V2 neural network. The improved YOLO-V2 neural network includes a residual block, a third convolution layer, and a fourth convolution layer. A first input of the residual block is connected to a first convolution layer of the improved YOLO-V2 neural network, and an output of the residual block is connected to a second convolution layer of the improved YOLO-V2 neural network. Here, the residual block is configured to transmit, to the second convolution layer, a summation result corresponding to the first convolution layer. The third convolution layer and the fourth convolution layer are generated by decomposing a convolution layer of an original YOLO-V2 neural network.