Light Weight Landmark Detection Method
- Resource Type
- Conference
- Authors
- Lv, Ying; Kong, Debao; Zhao, Yang; Yang, Zhaokai; Xu, Mengjie
- Source
- 2021 China Automation Congress (CAC) Automation Congress (CAC), 2021 China. :2811-2815 Oct, 2021
- Subject
- Aerospace
Communication, Networking and Broadcast Technologies
Components, Circuits, Devices and Systems
Computing and Processing
Power, Energy and Industry Applications
Robotics and Control Systems
Signal Processing and Analysis
Transportation
Deep learning
Automation
Intelligent vehicles
Computational modeling
Neural networks
Filtering algorithms
Feature extraction
landmark detection
Light weight
Attention mechanism
YOLOv3
- Language
- ISSN
- 2688-0938
Detecting landmarks in images is an important and challenging task for intelligent driving. The difficulty of this task is to locate and classify the landmarks accurately under the limited resources of intelligent driving platform. In this paper, we propose a method of filter pruning for deep neural networks for the real-time landmark detection. The network is based on yolov3 and improved for feature extraction. Then we conduct filter pruning based on the attention mechanism, resulting in the final lightweight model. The experiment on landmark detection datasets shows that the proposed lightweight landmark detection method can achieve high performance in accuracy with fast speed.