license: cc-by-nc-sa-4.0
language:
- zh
size_categories:
- 1K<n<10K
Dataset Description
- Homepage: ChinaOpen homepage
- Paper: ChinaOpen: A Dataset for Open-World Multimodal Learning
- Point of Contact: Xirong Li
Dataset Summary
ChinaOpen is a dataset sourced from Bilibili, a popular Chinese video-sharing website, for open-world multimodal learning. While the state-of-the-art multimodal learning networks have shown impressive performance in automated video annotation and cross-modal video retrieval, their training and evaluation are primarily conducted on YouTube videos with English text. Their effectiveness on Chinese data remains to be verified. For a multi-faceted evaluation, we build ChinaOpen-1k, a manually labeled test set of 1k videos. Each test video is accompanied with a manually checked user title and a manually written caption. Besides, each video is manually tagged to describe objects / actions / scenes shown in the visual content. The original user tags are also manually checked.
Languages
It is a Chinese dataset. Moreover, with all the Chinese text translated into English, ChinaOpen-1k is also suited for evaluating models trained on English data.
Dataset Structure
All the files are put in a zip package.
├── ChinaOpen-1k
├── video01.mp4
├── video02.mp4
├── video03.mp4
├── [...]
└── ChinaOpen-1k-annotations.json
Data Instances
Please refer to https://ruc-aimc-lab.github.io/ChinaOpen/#examples