# Fast Vehicle Detection in Aerial Imagery

**Authors:** Jennifer Carlet, Bernard Abayowa  
**arXiv:** [1709.08666](https://arxiv.org/abs/1709.08666)  
**Date:** 2017-09-25  
**Summary page:** https://www.summarizepaper.com/en/paper/1709.08666v1/

## Key points

- Significant progress in real-time or near real-time object detectors
- Detectors designed for first-person view images, not effective for detecting vehicles in aerial imagery
- Some detectors developed for aerial imagery but have slow processing speeds or struggle with multi-scale imagery
- Proposed modification to YOLOv2 detector to enhance performance on aerial data
- Comparison of proposed detector to Faster RCNN on various aerial imagery datasets
- Proposed detector achieves near state-of-the-art performance at more than four times the speed of existing methods
- Implications of findings: fast and accurate vehicle detection in aerial imagery for surveillance systems, traffic monitoring, and disaster response planning
- Innovative approach to improving object detection in aerial imagery by adapting an existing detector and achieving remarkable performance gains

## Summary

In recent years, there has been significant progress in the development of real-time or near real-time object detectors. However, these detectors are primarily designed for first-person view images where the subject is large in the image and do not perform well in detecting vehicles in aerial imagery. To address this limitation, some detectors have been specifically developed for aerial imagery but suffer from slow processing speeds or struggle to handle multi-scale imagery effectively. This study proposes a modification to the popular YOLOv2 detector to enhance its performance on aerial data and compares it to Faster RCNN on various aerial imagery datasets. The results demonstrate that the proposed detector achieves near state-of-the-art performance while operating at more than four times the speed of existing methods. Jennifer Carlet and Bernard Abayowa highlight the importance of their findings in enabling fast and accurate vehicle detection in aerial imagery with significant implications for applications such as surveillance systems, traffic monitoring, and disaster response planning. Overall, this study presents an innovative approach to improving object detection in aerial imagery by adapting an existing detector and achieving remarkable performance gains.