CVPR 2024 Conference Papers: Advances in AI and Computer Vision

This article highlights eight research papers from KuaiShou accepted into the CVPR 2024 conference, focusing on AI-driven innovations in computer vision, video processing, and 3D generation. Each paper addresses specific challenges in these domains through novel methodologies.

Key topics include multi-dimensional human preference modeling for text-to-image generation, efficient video quality assessment using pre-trained models, and code-sharing initiatives for compressed video enhancement. The papers demonstrate advancements in perception-oriented interpolation, test-time energy adaptation, and cross-domain retrieval techniques.

Several papers provide open-source code repositories, enabling further research and application in AI and multimedia fields. The work emphasizes practical implementations and benchmark validations across diverse datasets.