HaineiFRDM Film Restoration Dataset
收藏资源简介:
HaineiFRDM电影修复数据集由同济大学与上海电影修复实验室联合构建,包含真实退化胶片和合成数据两种类型。数据集包含5分钟训练用24fps真实修复影片及10分钟2K分辨率测试影片,通过人工筛选像素匹配的高清帧并标注视频描述,同时采用缺陷模板融合和画质退化模拟技术生成合成数据。该数据集旨在解决开源电影修复方法缺乏高质量训练数据的问题,为高分辨率胶片修复任务提供基准测试支持。
The HaineiFRDM Film Restoration Dataset was jointly developed by Tongji University and the Shanghai Film Restoration Laboratory, encompassing two categories of data: real degraded film footage and synthetic data. The dataset includes a 5-minute 24fps real restored film for training and a 10-minute 2K-resolution test film. High-definition frames with pixel-level matches are manually screened, and video descriptions are annotated for the dataset. Meanwhile, synthetic data is generated using defect template fusion and image quality degradation simulation technologies. This dataset aims to address the shortage of high-quality training data for open-source film restoration methods, and provides benchmark support for high-resolution film restoration tasks.




