PETS 2021
Virtual - November 16-19 2021

In Conjunction with AVSS 2021

PETS 2021 dataset

The FOLDOUT dataset

PETS 2021 introduces to the community a new scientfic challenge, specifically the problem of fragmented occlusion in tracking/detection by the practical task of through foliage detection; PETS2021 focuses on detection/tracking/recognition in natural forestry environments instead of man-made environments as it is common which raises new challenges to existing methods; PETS provides new multimodal data e.g. RGB/thermal which is previously unpublished and which the community is currently lacking. The data colllection was performed as part of the EU FOLDOUT project. The dataset provided for this challenge contains a set of four single camera video sequences (mixture of thermal and visible). The sequences contain one or more persons moving through the scene. As detailed below, there exist different inherent challenges within the data including the degree of fragmented occlusion and distance of the person(s) from the camera.

PETS 2021 Dataset characteristics


Detection and Tracking

The datasets made available present  low-level challenges on person detection and tracking.

We have selected and annotated some specific sequences specifically covering these challenges:


Table 1: PETS2021 Challenge Sequences

Challenge 1

Challenge 2

  • person detection through foliage

  • long-term person tracking in the wild


    • Challenge2_SEQ1; RGB camera, 847 frames, single person

 

Download Information

To download the video sequences please fill out the following User Agreement. You will receive a link to download the dataset after the User Agreement has been registered.

Legal note: The video sequences are copyright FOLDOUT project and the University of Reading and permission is hereby granted for free download for the purposes of the PETS 2021 challenge and academic and industrial research. Where the data is disseminated (e.g. publications, presentations) the source should be acknowledged.