Reproduce “Exploring AIGC Video Quality: A Focus on Visual Harmony, Video-Text Consistency and Domain Distribution Gap” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Bowen Qu, Xiaoyu Liang, Shangkun Sun, Wei Gao
Researchers from Peking University and Peng Cheng Laboratory developed a comprehensive framework for assessing the quality of AI-generated videos. This framework integrates evaluation of visual harmony, video-text consistency using multimodal language models, and explicit awareness of generative model domain differences, securing third place in the NTIRE 2024 AIGC Video Quality challenge.
Key insights
- A multi-dimensional approach, accounting for visual harmony, video-text consistency, and generator domain distribution, is essential for robust AIGC video quality assessment.
- Explicit and implicit integration of textual prompts and generated captions is crucial for comprehensively evaluating the semantic fidelity of AIGC videos.
- Incorporating an auxiliary task to classify generative model sources helps the quality assessment model learn more discriminative features, improving overall performance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.