MomentSeeker: A Comprehensive Benchmark and A Strong Baseline For Moment Retrieval Within Long Videos

This repo contains the annotation data for the paper "MomentSeeker: A Comprehensive Benchmark and A Strong Baseline For Moment Retrieval Within Long Videos".

🔔 News:

• 🥳 2025/03/07: We have released the MomentSeeker Benchmark and Paper! 🔥

License

Our dataset is under the CC-BY-NC-SA-4.0 license.

⚠️ If you need to access and use our dataset, you must understand and agree: This dataset is for research purposes only and cannot be used for any commercial or other purposes. The user assumes all effects arising from any other use and dissemination.

We do not own the copyright of any raw video files. Currently, we provide video access to researchers under the condition of acknowledging the above license. For the video data used, we respect and acknowledge any copyrights of the video authors. Therefore, for the movies, TV series, documentaries, and cartoons used in the dataset, we have reduced the resolution, clipped the length, adjusted dimensions, etc. of the original videos to minimize the impact on the rights of the original works.

If the original authors of the related works still believe that the videos should be removed, please contact [email protected] or directly raise an issue.

Introduction

We present MomentSeeker, a comprehensive benchmark to evaluate retrieval models' performance in handling general long-video moment retrieval (LVMR) tasks. MomentSeeker offers three key advantages. First, it incorporates long videos of over 500 seconds on average, making it the first benchmark specialized for long-video moment retrieval. Second, it covers a wide range of task categories (including Moment Search, Caption Alignment, Image-conditioned Moment Search, and Video-conditioned Moment Search) and diverse application scenarios (e.g., sports, movies, cartoons, and ego), making it a comprehensive tool for assessing retrieval models' general LVMR performance. Additionally, the evaluation tasks are carefully curated through human annotation, ensuring the reliability of assessment. We further fine-tune an MLLM-based LVMR retriever on synthetic data, which demonstrates strong performance on our benchmark. The checkpoint will release soon.

🏆 Mini Leaderboard

Rank	Method	Backbone	# Params	CA	MS	IMS	VMS	Overall
1	V-Embedder	InternVideo2-Chat	8B	42.2	20.4	15.0	15.8	23.3
2	CoVR	BLIP-Large	588M	25.8	17.4	12.3	12.3	17.1
3	InternVideo2	ViT	1B	44.6	18.2	4.8	0.0	16.9
4	MM-Ret	CLIP-Base	149M	23.2	15.4	10.5	10.5	14.9
5	LanguageBind	CLIP-Large	428M	39.6	16.4	3.2	0.0	14.8
6	E5V	LLaVA-1.6	8.4B	25.8	16.8	6.2	5.2	13.5
7	UniIR	CLIP-Large	428M	25.0	15.2	6.4	0.0	10.9
8	MLLM2VEC	LLaVA-1.6	8.4B	6.4	6.2	3.0	3.0	4.7
9	MagicLens	CLIP-Large	428M	9.0	2.4	3.2	2.8	4.4

License

Our dataset is under the CC-BY-NC-SA-4.0 license.

⚠️ If you need to access and use our dataset, you must understand and agree: This dataset is for research purposes only and cannot be used for any commercial or other purposes. The user assumes all effects arising from any other use and dissemination.

We do not own the copyright of any raw video files. Currently, we provide video access to researchers under the condition of acknowledging the above license. For the video data used, we respect and acknowledge any copyrights of the video authors. Therefore, for the movies, TV series, documentaries, and cartoons used in the dataset, we have reduced the resolution, clipped the length, adjusted dimensions, etc. of the original videos to minimize the impact on the rights of the original works.

If the original authors of the related works still believe that the videos should be removed, please contact [email protected] or directly raise an issue.

Evaluation

Before you access our dataset, we kindly ask you to thoroughly read and understand the license outlined above. If you cannot agree to these terms, we request that you refrain from downloading our video data.

The JSON file provides candidate videos for each question. The candidates can be ranked, and metrics such as Recall@1 and MAP@5 can be computed accordingly.

Hosting and Maintenance

The annotation files will be permanently retained.

If some videos are requested to be removed, we will replace them with a set of video frames sparsely sampled from the video and adjusted in resolution. Since all the questions in MomentSeeker are only related to visual content and do not involve audio, this will not significantly affect the validity of MomentSeeker (most existing MLLMs also understand videos by frame extraction).

If even retaining the frame set is not allowed, we will still keep the relevant annotation files, and replace them with the meta-information of the video, or actively seek more reliable and risk-free video sources.

Citation

If you find this repository useful, please consider giving a star 🌟 and citation

@misc{yuan2025momentseekercomprehensivebenchmarkstrong,
      title={MomentSeeker: A Comprehensive Benchmark and A Strong Baseline For Moment Retrieval Within Long Videos}, 
      author={Huaying Yuan and Jian Ni and Yueze Wang and Junjie Zhou and Zhengyang Liang and Zheng Liu and Zhao Cao and Zhicheng Dou and Ji-Rong Wen},
      year={2025},
      eprint={2502.12558},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2502.12558}, 
}