PETS 2021
Virtual - November 16-19 2021

In Conjunction with AVSS 2021

PETS 2021

IEEE International Workshop on
Performance Evaluation of Tracking and Surveillance


Rationale

Cameras_comparison-1024x576-1

The 2021 International Workshop on Performance Evaluation of Tracking and Surveillance (PETS'21) continues the series of highly successful PETS workshops held for over twenty years (AVSS'12, WVM'13, AVSS'14, AVSS'15, CVPR'16, CVPR'17, to name some of the most recent). The goal of the PETS workshop has been to foster the emergence of methodology for performance evaluation of tracking and surveillance.  This includes development of performance evaluation metrics as well as a quantitative evaluation of tracking and surveillance results based on a common dataset. PETS 2021 is sponsored by the EU Project FOLDOUT.

PETS 2021 introduces a new and exciting surveillance challenge, through-foliage detection and tracking. Such an application is important for (green) border surveillance. The specific tasks addressed are (1) through-foliage detection and fragmented occlusion, and (2) long-term tracking in natural environments, both of which have been received relatively limited attention in the computer vision community. Solutions to solve these tasks are currently unavailable. The aim of this challenge is to raise awareness to these tasks in the vision community and to foster appropriate solutions. The overall border surveillance challenge has significant impact for border authorities worldwide for enhancing border security operations. A corresponding dataset is provided for this challenge which covers the two tasks described above.

Ultimately, the workshop challenge comes down first, to accurately detecting and/or tracking person(s) present in the surveillance scene. To this end, a range of methods could be applied from change detectors to object detectors to trackers. Understanding of the temporal history of the movement of the person(s) may help in accurate localisation and tracking, especially in partially or heavily occluded scenes. Secondly, the challenge recognises that simple bounding boxes may not always be appropriate for annotating the visual extent of the detected object (person). Hence this challenge also considers alternate object representational schemes.

Challenge 1 - target detection through foliage 
Sequence 1
Sequence 2
Sequence 3



Three different image sequences containing different examples of persons walking undergoing fragmented occlusion.


Challenge 2 - long-term tracking in the wild


Representative images from a single long sequence containing a person moving across the surveillance scence undergoing occasional occlusion.


The workshop aims to bring together researchers, practitioners and students from computer vision and surveillance-related fields to share knowledge on methodologies, features and results related to the evaluation, modelling and understanding of object detection and tracking from video analytics.

This is a half-day workshop organised in association with  IEEE AVSS 2021, and financially co-sponsored by the EU project FOLDOUT.

A performance assessment of the submitted results will be performed by the challenge organisers and a challenge paper written up which reports the submmited methods and results' evaluation. The paper will be published at AVSS 2021 as part of the conference proceedings and indexed through IEEE Xplore. All challenge participants will be included as co-authors on the published paper.


Call for submissions


The workshop invites submissions which:
    Process (i.e. apply one or more change/object detectors and/or trackers) on one or more of the provided video sequences and report results in a standardised format. The submission should include a short description of the methodology used and results submitted in XML format.
Click on the Authors Instructions to know the detailed description of the challenges set in PETS 2021 and click on the Datasets link to see detailed description of the datasets and how to download them.

Please prepare the submissions according the "Authors Guidelines" provided under the Authors Instructions.


Important Dates


September 15, 2021
Deadline for submission of results and methods' description
September 22, 2021
Notification of acceptance
November 19, 2021
Challenge event (online)

Organisers


Organising Committee
James Ferryman (University of Reading)
Roman Pflugfelder (Austrian Institute of Technology)
Georg Melzer-Venturi (EUTEMA)
Email: j.m.ferryman@reading.ac.uk

Sponsors


This workshop is supported by the project FOLDOUT. FOLDOUT has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No 787021.