EditBoard
收藏资源简介:
EditBoard是由香港中文大学(深圳)、南京大学和牛津大学联合创建的第一个全面评估文本到视频编辑模型的基准。该基准涵盖了四个维度的九个自动评估指标,旨在通过详细的模型性能评估,提供对每个模型优缺点的深入见解。数据集的创建过程结合了先进的视频编辑技术和生成模型评估方法,旨在解决当前视频编辑模型评估中存在的不足,特别是在编辑任务的多样性和评估的一致性方面。EditBoard的应用领域主要集中在视频编辑模型的评估和改进,旨在推动视频编辑技术的标准化和进一步发展。
EditBoard is the first comprehensive benchmark for evaluating text-to-video editing models, jointly developed by The Chinese University of Hong Kong, Shenzhen, Nanjing University, and the University of Oxford. This benchmark encompasses nine automatic evaluation metrics across four dimensions, aiming to deliver in-depth insights into the strengths and weaknesses of each model through detailed performance assessments. The construction of this dataset integrates cutting-edge video editing technologies and generative model evaluation methodologies, with the objective of addressing the existing limitations in current text-to-video editing model evaluations, particularly regarding the diversity of editing tasks and the consistency of evaluation standards. The main application scenarios of EditBoard focus on the evaluation and improvement of video editing models, with the goal of promoting the standardization and further development of video editing technologies.

- 1EditBoard: Towards A Comprehensive Evaluation Benchmark for Text-based Video Editing Models香港中文大学(深圳) · 2024年



