Pet Recognition Method in Mobile Community Based on Convolutional Neural Network

Abstract

The invention discloses a mobile terminal pet identification method based on a convolutional neural network. The method includes: constructing a lightweight convolutional neural network, using a pre-trained network model training method on the server side to train the convolutional neural network, deploying the trained convolutional neural network model on the mobile terminal device through a mobile terminal deep learning framework, and passing The mobile device collects images and performs target recognition and risk judgment. The disclosure of the present invention realizes the detection of community pet types on the mobile terminal equipment, and residents can stay away from dangerous pets through the identification results, effectively avoiding the occurrence of pets hurting people in the community; in addition, the lightweight The quantized convolutional network does not have high requirements on the hardware environment, and can adapt to most mobile operating environments while ensuring recognition accuracy.

Description

technical field [0001] The invention relates to the technical field of image recognition, and in particular, designs a mobile terminal community pet recognition method based on a convolutional neural network. Background technique [0002] With the continuous improvement of people's living standards, more and more people like to keep one or several pets to accompany their lives. Especially in cities, not only the number of pets is increasing, but also the types are also varied. This has led to a series of problems: among many pets, there are large dogs or fierce dogs. The former are larger and their sudden actions are beyond the controllable range of manpower, while the latter are inherently aggressive and prone to injury. In addition, some pets with small size also have a strong sense of precaution and territorial awareness, and they are also aggressive, and pets may carry viruses or bacteria, which will cause serious consequences if they hurt people. . [0003] Although s...

AI Technical Summary

Problems solved by technology

This has led to a series of problems: Among the many pets, there are many large dogs or fierce dogs. The former are large in size and their sudden actions are beyond the controllable range of manpower. The latter are naturally aggressive and prone to injuries. In addition, some small pets also have a strong sense of defense and territoriality, and are also aggressive, and pets may carry viruses or bacteria, and once they hurt people, it will cause serious consequences

However, there are still problems with the current algorithms: most of the current algorithms are dedicated to improving the detection accuracy and real-time performance to the extreme, so the scale of the network model and the amount of calculation are huge, which is suitable for the GPU server hardware operating environment, but for mobile devices installed Or the embedded video surveillance system in the community, its real-time performance still cannot meet the actual application requirements

Images