Vehicle component detection model compression method and system

Abstract

The invention discloses a vehicle component detection model compression method and system based on Faster R-CNN, and aims at reducing the occupancy rate of the vehicle component detection model in thememory space and reducing the model detection time. The method comprises the steps that the basic vehicle component detection model is trained, model channel pruning is performed and the model is quantified. In the compression process of the vehicle component detection model, the basic vehicle component detection model M1 is trained firstly, then the redundant channels in the M1 are identified and deleted so as to obtain M2, and finally parameter quantification of M2 is performed so as to obtain M3. The problem of low speed of vehicle component detection by using the Faster R-CNN algorithm can be effectively solved and the occupancy rate of the model in the memory space can be reduced.

Description

technical field [0001] The invention belongs to the field of pattern recognition, and more specifically relates to a method and system for compressing a vehicle component detection model based on Faster R-CNN, which is used to reduce the memory space occupation rate of the vehicle component detection model and shorten the detection time. Background technique [0002] Intelligent transportation systems can be used to help solve social problems such as traffic accidents, traffic congestion, and car theft. The detection of vehicle components such as license plates, car logos, lights, and windshields is a very basic and important part of the intelligent transportation system. ring. Because they contain vehicle appearance and structure information, they can fully reflect the attributes of the vehicle, and are stable and difficult to change. If these vehicle parts can be detected accurately and quickly, the function of the entire intelligent transportation system will be effectiv...

AI Technical Summary

Problems solved by technology

[0004] In view of the above defects or improvement needs of the prior art, the present invention provides a vehicle component detection model compression method and system, thereby solving the problem of detecting vehicle component speed when using Faster R-CNN as the basic detection algorithm to detect vehicle components Technical issues with slowness and high memory space usage

Images