Preparing Video Retrieval Datasets

Introduction

@inproceedings{xu2016msr,
      title={Msr-vtt: A large video description dataset for bridging video and language},
      author={Xu, Jun and Mei, Tao and Yao, Ting and Rui, Yong},
      booktitle={CVPR},
      pages={5288--5296},
      year={2016}
}

@inproceedings{chen2011collecting,
  title={Collecting highly parallel data for paraphrase evaluation},
  author={Chen, David and Dolan, William B},
  booktitle={ACL},
  pages={190--200},
  year={2011}
}

Before we start, please make sure that the directory is located at $MMACTION2/tools/data/video_retrieval/.

Preparing MSRVTT dataset

For basic dataset information, you can refer to the MSRVTT dataset website. Run the following command to prepare the MSRVTT dataset:

bash prepare_msrvtt.sh

After preparation, the folder structure will look like:

mmaction2
├── mmaction
├── tools
├── configs
├── data
│   ├── video_retrieval
│   │   └── msrvtt
│   │       ├── train_9k.json
│   │       ├── train_7k.json
│   │       ├── test_JSFUSION.json
│   │       └─── videos
│   │           ├── video0.mp4
│   │           ├── video1.mp4
│   │           ├── ...
│   │           └── video9999.mp4

Preparing MSVD dataset

For basic dataset information, you can refer to the MSVD dataset website. Run the following command to prepare the MSVD dataset:

bash prepare_msvd.sh

After preparation, the folder structure will look like:

mmaction2
├── mmaction
├── tools
├── configs
├── data
│   ├── video_retrieval
│   │   └── msrvd
│   │       ├── train.json
│   │       ├── test.json
│   │       ├── val.json
│   │       └─── videos
│   │           ├── xxx.avi
│   │           ├── xxx.avi
│   │           ├── ...
│   │           └── xxx.avi