BirdsEyeView: Aerial View Dataset for Object Classification and Detection

In recent years, deep learning based computer vision technology has progressed rapidly thanks to the significant increases in computing power and high-quality datasets. In this article, we present an aerial view image and video dataset dedicated to facilitating vision applications on the UAV platform, such as object detection, classification and tracking. The dataset consists of 5,000 images, each of which is carefully annotated according to the guidelines of the PASCAL VOC. The dataset is designed to cover diverse real-life scenes with aerial view angles which is different from other datasets. Such kind of specific dataset will be of great importance in developing and testing deep learning algorithms for UAV applications. Moreover, the dataset can serve as a benchmark to evaluate UAV visual solutions.

I. Introduction

In the past few years, deep learning based computer vision technology has attracted tremendous attention due to its state-of-the-art performance on a wide range of visual applications. Both academia and industry have made significant progress in several core techniques such as object detection and object tracking. From these techniques, various applications are enabled such as surveillance, resource monitoring, and wilderness search and rescue.