MS-CADNet: A Multi-Scale Context Attention Network for Efficient Object Detection in UAV Imagery
Abstract
With the rapid advancement of unmanned aerial vehicle (UAV) technology, there is a need for lightweight and accurate object detection on resource-constrained platforms. This paper proposes MS-CADNet, an anchor-free network for small object detection in aerial imagery. It uses a MobileNetV3-Small backbone and a two-branch gated Context Attention Module (CAM) to enhance feature quality. On the VisDrone-DET benchmark, it achieves 31.2% mAP, surpassing YOLOv8-Small and CEASC. The model attains 19.2% AP for small objects with only 3.1M parameters and 5.4 GFLOPs, making it suitable for real-time UAV deployment.
Fields of Study & Subjects
Advanced Neural Network Applications
UAV Applications and Optimization
Domain Adaptation and Few-Shot Learning
Publication details
- Journal / source
- ICCK Transactions on Sensing Communication and Control
- Publisher
- Institute of Central Computation and Knowledge Inc.
- Volume
- 3
- Issue
- 2
- Pages
- 64–75
- Publication date
- 23 April 2026