Research on Few-Shot Object Counting Model for Smart Parking in Future Communities
Ensuring a more user-friendly and intelligent provision of parking guidance services to residents is fundamental to future community development.This paper presents an object-counting-based approach for accurate and real-time detection of available parking spaces,offering technological support for this service.The model employs a convolutional-neural-network-based density estimation technique to count vehicles and available parking spaces.In addition to providing parking guidance for residents,it also serves as a source of traffic flow data for community management and facilitates macro-level data aggregation for multi-community parking space sharing;meanwhile,we perform data augmentation for the existing object counting method,and the experimental results show that the MAE of the model is improved by 15.7%and the RMSE is improved by 25.81%after data augmentation.Furthermore,this model demonstrats excellent versatility and can be applied to various other domains.
future communityobject countingsmart parkingfew-shot counting model
NETL
NSTL
万方数据
维普
