This is an official pytorch implementation of our CVPR 2022 paper 3MASSIV: Multilingual, Multimodal and Multi-Aspect dataset of Social Media Short Videos. 3MASSIV is a human-annotated, multimodal and multilingual dataset containing short videos uploaded to a leading short-video social media platform - MOJ by ShareChat.
In this repository, we provide the dataset and codebase for exploring our work.
Please visit the Project Page for more details about the dataset.
3MASSIV dataset is available for non-commercial research purposes only.
The training, validation and testing splits are available in the data folder.
For downloading the videos, please sign the agreement.
Please ask your supervisor/advisor to sign the agreement appropriately and then send the scanned version to Vikram Gupta [vikramgupta(at)sharechat(dot)co] and Prima [prima(dot)indic(at)gmail(dot)com]
After verifying your request, we will contact you with the password to download the videos.
For reproducing the results reported in the paper, please refer to the README.md and training scripts present in the code folder.
If you find 3MASSIV useful in your research, please use the following BibTeX entry for citation.
@article{gupta20223massiv,
title={3MASSIV: Multilingual, Multimodal and Multi-Aspect dataset of Social Media Short Videos},
author={Gupta, Vikram and Mittal, Trisha and Mathur, Puneet and Mishra, Vaibhav and Maheshwari, Mayank and Bera, Aniket and Mukherjee, Debdoot and Manocha, Dinesh},
journal={arXiv preprint arXiv:2203.14456},
year={2022}
}