UNN-6: An Activities of Daily Life Dataset
for Fall Detection and Recognition

Ryan Cameron, Zheming Zuo, Graham Sexton, and Longzhi Yang
Department of Computer and Information Sciences, Northumbria University, Newcatle upon Tyne, U.K.

Introduction

The UNN-6 includes a colour (RGB) clip set and an infrared (IR) clip set, with sample frames shwon above. Each set contains 36 clips (i.e. 6 videos per class) that were recorded using the Microsoft® Kinect™ v2 in an indoor environment with varying lighting conditions and dynamic backgrounds (e.g. TV is playing in the background). For simulating real-world CCTV system and improving the computational efficiency , the resolution for all video clips are uniformly resized from 1920×1080 to 320×240. Each video clip in both sets are cropped into 150 frames with 25 frame rate applied, i.e. each video lasts for 6 seconds as a fall or similar action always happens within 6 seconds. In addition, the generated IR clips do not include any depth information, as depth cameras are still costly and not widely deployed for digital healthcare.

Download

  • UNN-6 Dataset
    • contains both Colour Set and InfraRed Set (41 MB)
    • [Google Drive]
  • UNN-6 Colour Set
  • UNN-6 InfraRed Set

Citation


If you use this datset in your research, please refer to the following paper: 
@InProceedings{ryan17FallDetection,
author={Cameron, Ryan and Zuo, Zheming and Sexton, Graham and Yang, Longzhi",
title={A Fall Detection/Recognition System and an Empirical Study of Gradient-Based Feature Extraction Approaches},
booktitle={Advances in Computational Intelligence Systems},
year={2018},
publisher={Springer International Publishing},
address={Cham},
pages={276--289},
isbn={978-3-319-66939-7},}

For more experimental results on the UNN-6 dataset, please refer to the following work: 
@INPROCEEDINGS{zuo18SstVladSstFv,
    author    = {Zheming Zuo and Daniel Organisciak and Hubert P. H. Shum and Longzhi Yang},
    title     = {Saliency-Informed Spatio-Temporal Vector of Locally Aggregated Descriptors and Fisher Vectors for Visual Action Recognition},
    booktitle = {2018 BMVA British Machine Vision Conference (BMVC)},
    pages     = {321.1--321.11},
    year      = {2018}}