UAV123
2D Box
Object Tracking
|Common
|...
License: Unknown

Overview

A Benchmark and Simulator for UAV Tracking (Dataset)

Video captured from low-altitude UAVs is inherently different from video in popular tracking datasets like OTB50, OTB100, VOT2014, VOT2015, TC128, and ALOV300++. Therefore, we propose a new dataset (UAV123) with sequences from an aerial viewpoint, a subset of which is meant for long-term aerial tracking (UAV20L).

Data Collection

Our new UAV123 dataset contains a total of 123 video sequences and more than 110K frames making it the second-largest object tracking dataset after ALOV300++.

Data Annotation

All sequences are fully annotated with upright bounding boxes. The dataset can easily be integrated with the visual tracker benchmark. It includes all bounding box and attribute annotations for the UAV dataset.
For more details, this document containing all annotation details [pdf].

Instruction

The setup for the dataset can be found in configSeqs.m.
The bounding box and attribute annotation can be found in the folder anno.
For more details, please read ReadMe.txt after downloading.

Citation

@Inbook{Mueller2016,
author="Mueller, Matthias and Smith, Neil and Ghanem, Bernard",
editor="Leibe, Bastian and Matas, Jiri and Sebe, Nicu and Welling, Max",
title="A Benchmark and Simulator for UAV Tracking",
bookTitle="Computer Vision -- ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11--14, 2016, Proceedings, Part I",
year="2016",
publisher="Springer International Publishing",
address="Cham",
pages="445--461",
isbn="978-3-319-46448-0",
doi="10.1007/978-3-319-46448-0_27",
url="http://dx.doi.org/10.1007/978-3-319-46448-0_27"
}
Data Summary
Type
Image,
Amount
--
Size
--
Provided by
IVUL
At IVUL, we focus on interesting research problems that arise in computer vision, including activity recognition/detection, robust representations of objects for tracking and recognition, scene understanding from 3D data, image annotation, etc. Since our goal is to make sense of images and videos especially at large-scales, we also often end up developing new machine learning and optimization methods to help us achieve this goal.
Issue
Start Building AI Now