Tasks | Dataset | Leaderboard | Paper

This repository contains information about the general multi-modal evaluation benchmark GEM, which is composed of GEM-I for image tasks spans 20 languages and GEM-V for video tasks spans 30 languages. The current version of GEM is composed of 8 tasks. For each task, training and validation set are provided. GEM is not only the largest vision-language dataset covering image-language tasks and video-language tasks at the same time, but also labeled in multiple languages.

GEM-I contains 1.2 million {Query, Image, Title} triplets in 20 different languages for text-to-image retrieval and image captioning tasks. The statistics can be found in below table.

GEM-V contains 99K {Query, Video, Title} triplets in 30 languages for text-to-video retrieval and video captioning tasks. The statistics can be found in below table.

The 8 tasks in GEM can be categorized into 4 groups: image retrieval tasks, image captioning tasks, video retrieval tasks and video captioning tasks.

Within each language, we use query to retrieve images, and the evaluation metric is mean-Recall (arithmetic mean of Recall@K for K in {1, 5, 10}).

Within each language, we use query to retrieve images with title as context, and the evaluation metric is mean-Recall (arithmetic mean of Recall@K for K ∈ {1, 5, 10}).

We use image to generate caption text, and use ROUGE-L as the evaluation metric.

Within each language, we use query to retrieve videos, and the evaluation metric is mean-Recall (arithmetic mean of Recall@K for K in {1, 5, 10}).

Within each language, we use query to retrieve videos with title as context, and the evaluation metric is mean-Recall (arithmetic mean of Recall@K for K in {1, 5, 10}).

We use video to generate caption text, and use ROUGE-L as the evaluation metric.

We use Title to generate caption text, and use ROUGE-L as the evaluation metric.

We use video and title to generate caption text, and use ROUGE-L as the evaluation metric.

In order to use our dataset, please navigate to GEM Leaderboard and agree to our terms of service. After you do so a download link will be made available.

To submit your predictions for evaluation, please create a single folder which contains the 8 sub-folders named after each task. Inside each folder, create one prediction file for each language and name the file using the following format: {language}.prediction where {language} is the 2 character ISO 639-1 code. Please email your submission. We will reply with your model performance.

For self-evaluation on dev set, you can refer to below evaluation scripts:

1. Image Retrieval: ./evaluation/metric-retrieval.py
2. Image Captioning: https://github.com/tylin/coco-caption
3. Video Retrieval: ./evaluation/metric-retrieval.py
4. Video Captioning: https://github.com/Maluuba/nlg-eval

To evaluate your model's performance, we will compare your prediction files with the ground truth files. We are keeping our evaluation data held out but we ask all models first evaluate performance on the development portion of the dataset before submitting their predictions for the evaluation dataset.

The detailed format of each task is at Evaluation ReadMe.

If you use our benchmark or dataset, please cite our paper \cite{lin2021gem}.

@inproceedings{lin2021gem,
    title = "{GEM}: A General Evaluation Benchmark for Multimodal Tasks",
    author = "Lin Su and Nan Duan and Edward Cui and Lei Ji and Chenfei Wu and Huaishao Luo and Yongfei Liu and Ming Zhong and Taroon Bharti and Arun Sacheti",
    booktitle = "Findings of the Association for Computational Linguistics",
    month = aug,
    year = "2021",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "",
    doi = "",
    pages = "",
    abstract = "",
}

This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.opensource.microsoft.com.

When you submit a pull request, a CLA bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., status check, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.

This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact [email protected] with any additional questions or comments.

This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow Microsoft's Trademark & Brand Guidelines. Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos are subject to those third-party's policies.