VideoScore: Building Automatic Metrics to Simulate Fine-grained Human Feedback for Video Generation

TL;DR

This paper introduces VideoScore, an automatic video quality assessment metric trained on a large-scale human-annotated dataset, achieving high correlation with human judgments and enabling better evaluation and improvement of video generation models.

Contribution

The paper presents VideoFeedback, the first large-scale human-annotated dataset for video quality, and develops VideoScore, a new metric that outperforms existing methods in correlating with human judgments.

Findings

VideoScore achieves 77.1 Spearman correlation with human scores.

VideoScore outperforms prior metrics by about 50 points.

VideoScore consistently correlates better with human judges across multiple benchmarks.

Abstract

The recent years have witnessed great advances in video generation. However, the development of automatic video metrics is lagging significantly behind. None of the existing metric is able to provide reliable scores over generated videos. The main barrier is the lack of large-scale human-annotated dataset. In this paper, we release VideoFeedback, the first large-scale dataset containing human-provided multi-aspect score over 37.6K synthesized videos from 11 existing video generative models. We train VideoScore (initialized from Mantis) based on VideoFeedback to enable automatic video quality assessment. Experiments show that the Spearman correlation between VideoScore and humans can reach 77.1 on VideoFeedback-test, beating the prior best metrics by about 50 points. Further result on other held-out EvalCrafter, GenAI-Bench, and VBench show that VideoScore has consistently much higher correlation with human judges than other metrics. Due to these results, we believe VideoScore can serve as a great proxy for human raters to (1) rate different video models to track progress (2) simulate fine-grained human feedback in Reinforcement Learning with Human Feedback (RLHF) to improve current video generation models.