Paper Info
Title
DAFV: A Unified and Real-Time Framework of Joint Detection and Attributes Recognition for Fast Vehicles
Abstract
In the past decade, with the development of computing equipment and CNN, target detection has made great progress, which has promoted the development of specific target detection. The purpose of vehicle detection is not only to extract the vehicle from a large number of traffic surveillance cameras, but also for some follow-up research, such as the structured storage of vehicle information, which needs to quickly identify the attributes of the vehicle. Based on those demands, we propose a method of joint Detection and Attributes recognition for Fast Vehicles (DAFV). Firstly, we present Feature Rapidly Extract Module (FREM), which is to quickly shrink the feature map size and enhance the runtime efficiency. Secondly, we present Feature Refinement Module (FRM) to increase feature utilization rate and improve the performance. Lastly, we present the Cross-Stage and Multi-Scale (CS-MS) Module to optimize scale-invariant design. Related experiments based on UA-DETRAC dataset proves that DAFV is a feasible and effective method. The DAFV is fast and the speed does not change with the number of vehicles. For 416 x 416 pictures, DAFV can reach 53 FPS with only 775 Mib GPU memory, which can meet the needs of real-time applications.
Year
DOI
Venue
2021
10.1007/978-3-030-86130-8_28
WIRELESS ALGORITHMS, SYSTEMS, AND APPLICATIONS, WASA 2021, PT II
Keywords
DocType
Volume
Vehicle detection, Attributes recognition, Real-time
Conference
12938
ISSN
Citations 
PageRank 
0302-9743
0
0.34
References 
Authors
0
5
Name
Order
Citations
PageRank
Yifan Chang100.34
Chao Li2253.52
Zhiqiang Li300.34
Zhiguo Wang400.34
Guangqiang Yin525.79