YFCC100M video metadata database
Introduction
The Yahoo Flickr Creative Commons 100 Million dataset (YFCC100M) is one of the largest public multimedia databases which contains 99.2 million images and 0.8 million videos. Although its videos take only a small part of the dataset in comparison to its images, the video part is still one of the largest publicly available video databases. However, some essential video metadata and features are missing in the original database. Therefore, we obtained several metadata and features from the videos and release the database here for the research community.
Description
Metadata included
This database includes several essential spatial, temporal, and content features of videos in the YFCC100M dataset. Following features and metadata are included in this database.
- Spatial features: aspect ratio, orientation
- Temporal features: duration (sec.), number of frames, frames per second
- Content features: 100-bin hue histogram, spatial information (SI), temporal information (TI)
For more information, please refer to our paper which is mentioned below.
Structure
The database is in Tab-Separated Values (TSV) format with LF (Unix) line endings. Each line contains following tab-separated fields of each video:
- Video hash in the original YFCC100M dataset
- Aspect ratio
- Orientation (P = portrait, L = landscape, S = 1:1)
- Duration (sec.)
- Number of frames
- Frame rate (FPS)
- Spatial information (SI)
- Temporal information (TI)
- 1st bin of the 100-bin hue histogram
- 2nd bin of the 100-bin hue histogram
- ...
- 100th bin of the 100-bin hue histogram
License note
The database is available under Creative Commons Attribution 3.0 license. Please cite following paper when you use this database:
- J.-H. Choi and J.-S. Lee, "Analysis of spatial, temporal, and content characteristics of videos in the YFCC100M dataset," ACM Multimedia (MM), Multimedia COMMONS Workshop, Oct. 2016
Detail
Download
- yfcc100m_videos.zip (219MB)
- SHA1: cc019c4119a06bc8a7197a42d335d94c0ec21244
- Last updated: Oct. 15, 2016
- Oct. 15, 2016: initial release
Contact
- Jun-Ho Choi: idearibosome at yonsei.ac.kr
- Jong-Seok Lee: jong-seok.lee at yonsei.ac.kr