A Database for Multi-Modal Short Video Quality Assessment

The short video has gained increasing attention in information sharing and commercial promotions due to the fast development of social platforms. Accompanying, it introduces great requirements for assessing the quality of short videos for efficient information acquirement and propagation. However, existing video quality assessment researches focus on assessing video content with five rating scores, limiting the assessment to a one-dimension and simplified criterion. In this paper, we establish a novel database dubbed MMSVD-Douyin for assessing multi-modal short video quality under consideration of three evaluation criteria. It includes 4,684 short videos, three kinds of modalities, six kinds of data formats, and three assessment criteria. To conduct the short video quality assessment, we set up an all-around multi-modal short video quality assessment benchmark (MulSVQA) that dynamically fuses representations from three modalities and produces numbers of "likes", "shares" and "comments" of short videos.

1. INTRODUCTION

As portable devices and social media applications become widely accessible, users are getting used to capturing information with short videos, which possess characteristics of multiple modalities, limited-time series, and multi-facet assessments. A short video always consists of multiple modalities including image (cover image), video (video content), and text language (video title, author information, etc.), and be tagged as "clicks", "likes", "shares", "comments", etc. According to statistics, the total number of monthly active users existing in TikTok and Douyin (two popular video-sharing and social media platforms) achieves 1.2 billion [1] and 934 million [2], respectively. With a huge user community, short videos have accounted for a large portion of internet traffic and become a commercially important carrier, thus bringing about contributions to commercial promotions and gains. A practical issue arising from short videos is assessing values that may be potentially produced, i.e., Short Video Quality Assessment (SVQA).