HD-dSEQUA Video Sequence Database

The HD-dSEQUA Video Sequence Database consists of raw video files and a data file with quality indicators.

This dataset was created to support research on video compression and quality assessment. It contains 131 diverse video sequences that vary in spatial, temporal, and semantic characteristics. The materials come from multiple sources such as CableLabs, Netflix, SJTU Media Lab, Xiph.org, Ultra Video Group, and Twitch streams, ensuring broad coverage of both professional and user-generated content.

The videos are stored in the Y4M format (uncompressed), mostly in Full HD resolution (1920×1080), with additional sequences in 1920×804 and Ultra HD (3840×2160). Frame rates range from 24 FPS to 60 FPS, with most videos lasting between 10 and 20 seconds.

The CSV file contains a table with video identifiers, names, and values of 10 no-reference Video Quality Indicators (VQIs), including measures of spatial and temporal activity, blur, noise, exposure, contrast, blockiness, block loss, interlacing artifacts, and UGC classification. These indicators enable quick comparison and benchmarking of video material quality.

Both the videos and the data file are available at: qoe.agh.edu.pl/hd-dsequa.

If you use these video sequences, please reference this web page as well as our paper, where a detailed description of the database has been published:

Korus, Filip; Juszka, Dawid; Grega, Michał; Leszczuk, Mikołaj; Suchoń, Michał; Dutta, Avrajyoti; Wanat, Dominika

HD-dSEQUA: high definition diverse video sequence dataset for quality assessment with indicators Journal Article

In: Przegąd Telekomunikacyjny — Wiadomości Telekomunikacyjne, vol. 98, no. 4, pp. 451-454, 2025.

BibTeX

en_USEnglish