Rapid image recognition accelerator design method based on convolutional neural network

Abstract

The invention relates to a rapid image recognition accelerator design method based on a convolutional neural network, and belongs to the technical field of image recognition. The method comprises thefollowing steps: 1) based on an Internet of Things terminal combining ARM and FPGA, configuring parameters of a camera at an ARM end, and processing acquired image data and weight data; 2) designing an assembly line processing scheme combining software and hardware, adopting an operation strategy combining image block parallelism, input channel parallelism and output channel parallelism, and establishing a model of terminal resources and identification time based on the strategy; and 3) acquiring an optimal image block size and convolution parallel parameters according to the solving model, and constructing a convolutional neural network model at the FPGA end to identify the image. According to the invention, on-chip resources can be fully utilized on a resource-limited Internet of Thingsterminal, and the image identification speed is effectively improved while the resource utilization rate is improved.

Description

technical field [0001] The invention belongs to the technical field of image recognition, and relates to a design method of a fast image recognition accelerator based on a convolutional neural network. Background technique [0002] With the development of the Internet of Things and artificial intelligence, the requirements for target detection and face recognition are getting higher and higher in scenarios such as smart transportation and video surveillance in the field of Internet of Things. Convolutional Neural Network (CNN) is used as an image recognition Commonly used algorithms, which play a vital role in scenarios such as object detection in the Internet of Things field. However, convolutional neural network algorithms require a lot of resource consumption, while IoT terminal devices are often resource-intensive, and have higher requirements for cost, power consumption, and real-time performance. How to apply convolutional neural network algorithms to IoT terminals wi...

AI Technical Summary

Problems solved by technology

[0003] FPGA provides a feasible solution for the application of deep learning algorithms in the field of Internet of Things terminals. Although good computing performance is achieved on FPGA-based Internet of Things terminals, there are fewer considerations for resource-constrained Internet of Things terminals and there are still room for improvement

Images