Deep learning target detection system based on server-embedded cooperation

Abstract

The invention discloses a deep learning target detection system based on server-embedded cooperation. A server side comprises a knowledge base, a training model, a test result statistical analysis module and a calculation resource monitoring module. The knowledge base comprises a data management module, the training model comprises a deep learning network training module and a model compression module, the test result statistical analysis comprises a model test module, and the calculation resource monitoring module is used for monitoring CPU and GPU resources of the system in real time, so that a user can reasonably utilize the calculation resources according to actual conditions; the embedded end comprises a main controller ARM and a coprocessor FPGA. The main controller ARM is responsible for external image input, image preprocessing, an NMS algorithm, superposition detection information and an image output task. The coprocessor FPGA is responsible for acceleration of a convolution layer, a shortcut layer and an up-sampling layer in deep network reasoning. According to the method, the problem that the existing deep learning network model is quickly deployed from training of a server side to an embedded side is solved.

Description

technical field [0001] This application designs the technical field of server-side training and embedded-side deployment, and in particular relates to a deep learning target detection system based on server-embedded collaboration. Background technique [0002] Deep learning is a data-driven machine learning algorithm with super feature self-extraction capabilities. Its performance in image target detection, especially in multi-type, multi-scale, and multi-angle target detection, far exceeds traditional machine learning. method. With the advent of the Internet of Things era and the rise of edge computing, the deep learning object detection network is gradually developing from the server side to the embedded side, realizing the interconnection and perception of all things. [0003] The essence of deep learning is a signal processing system formed by stacking multiple layers of artificial neural networks under the support of big data, which has the characteristics of large num...

AI Technical Summary

Problems solved by technology

[0005] Currently, server-side deep learning network model training, model compression, and embedded platform acceleration are independent of each other. There is a lack of a target detection system that can open up the data paths of the three, and each module is not packaged into an easy-to-operate software interface, which increases the need for rapid application. In order to reduce the threshold of deep learning model training, compression and embedded deployment, a "server-embedded" collaborative deep learning target detection system is needed, which uses interface visualization to reduce the threshold of deep learning target recognition network learning, helping non-human Deep learning personnel can quickly get started with the application, so as to achieve the purpose of fast implementation of deep learning target detection applications

Images