Results of the 2024 Video Browser Showdown
Luca Rossetto, Klaus Schoeffmann, Cathal Gurrin, Jakub Lokoč, Werner Bailer
TL;DR
The paper presents a concise results-focused report on VBS 2024, the 13th edition of the interactive video retrieval competition. It analyzes performance across four task types (AVS, KISV, KIST, QAS) over three challenging datasets, with QAS introduced this year, and compares expert and novice user sessions to produce a final per-team score based on the best-performing participant in each session. The study leverages the DRES-TOMM evaluation server for coordination and data provenance and includes live judging for non-automatically-verified submissions. Key contributions include the first incorporation of QAS tasks, detailed per-task and per-session scoring, and dynamic rankings over time, offering insights into system effectiveness and browsing efficiency for interactive video retrieval. The findings inform future VBS iterations and contribute to benchmarks for real-time, user-driven video search systems in heterogeneous content domains.
Abstract
This report presents the results of the 13th Video Browser Showdown, held at the 2024 International Conference on Multimedia Modeling on the 29th of January 2024 in Amsterdam, the Netherlands.