A unified toolkit for visual object tracking and segmentation evaluation

Čehovin Zajc, Luka (author)

We present a toolkit for evaluating and analyzing visual object tracking and segmentation methods. In contrast to most existing tools, the proposed toolkit introduces a unified representation of these tasks, enabling tracking of one or more objects with different object descriptions, from points to segmentation masks. Moreover, it is organized in a modular manner, supporting multiple integration types, various experiment types, and comprehensive reporting. The toolkit’s purpose aligns with the goals of the tracking community by supporting evaluation for all VOT and VOTS challenges, as well as for several other video-tracking or segmentation datasets. It is also extensible in almost every direction (data, methods, experiment design). This paper illustrates the toolkit’s architecture and usage patterns, and discusses its role in supporting the community’s transition towards unified video understanding.

• Language: English
• Keywords: tracking, video segmentation, performance evaluation, Python
• Work type: Article
• Typology: 1.01 - Original Scientific Article
• Organization: FRI - Faculty of Computer and Information Science
• Publication status: Published
• Publication version: Version of Record
• Year: 2027
• Number of pages: 6 str.
• Numbering: Vol. 709, art. 135295
• PID: 20.500.12556/RUL-190183
• UDC: 004.93:004.42
• ISSN on article: 0925-2312
• DOI: 10.1016/j.neucom.2026.135295
• COBISS.SI-ID: 302073859
• Publication date in RUL: 23.10.2026

Record is a part of a journal

• Title: Neurocomputing
• Shortened title: Neurocomputing
• Publisher: Elsevier
• ISSN: 0925-2312
• COBISS.SI-ID: 172315

Licences

• License: CC BY 4.0, Creative Commons Attribution 4.0 International
• Link: http://creativecommons.org/licenses/by/4.0/
• Description: This is the standard Creative Commons license that gives others maximum freedom to do what they want with the work as long as they credit the author.

Secondary language

• Language: Slovenian
• Keywords: sledenje, video segmentacija, ocenjevanje performans, Python

Projects

• Funder: ARIS - Slovenian Research and Innovation Agency
• Project number: Z2-1866
• Name: Globoko generativno modeliranje izgleda v vizualnem sledenju

• Funder: ARIS - Slovenian Research and Innovation Agency
• Project number: J2-60045
• Name: RoDEO: Robustno globoko učenje za opazovanje Zemlje