Comparison of Deep Learning-Based Semantic Segmentation Models for Unmanned Aerial Vehicle Images

Abstract

Unmanned aerial vehicles (UAVs) have been used for a variety of tasks, including transporting food, medical supplies, packages, and other items. Semantic segmentation allows us to understand urban scenes, which is important for improving the safety of autonomous UAVs. In this paper, we investigated several deep learning-based models for segmentation in aerial images. The models are built based on FeN, U-Net, and DeepLab architectures. We trained and evaluated models using a publicly available dataset. The experimental results show that all models have high potential with a small number of training samples. We also compared the results and provided possible suggestions for further work.