Towards Data-Efficient Video Pre-training with Frozen Image Foundation Models
Quick Take
This study explores using frozen image models for efficient video pre-training focused on temporal reasoning.
Key Points
- Video models typically require extensive data and compute resources.
- Frozen image models can serve as spatial encoders.
- Initial results show strong temporal performance with reduced video data.
📖 Reader Mode
~2 min readAbstract:Video foundation models achieve strong performance across many video understanding tasks, but typically require large-scale pre-training on massive video datasets, resulting in substantial data and compute costs. In contrast, modern image foundation models already provide powerful spatial representations. This raises an important question: can competitive video models be built by reusing these spatial representations and pre-training only for temporal reasoning? We take initial steps toward exploring a lightweight training paradigm that freezes a pre-trained image foundation model and trains only a recurrent temporal module to process streaming video. By reusing an image foundation model as a spatial encoder, this approach could significantly reduce the amount of video data and compute required compared to end-to-end video pre-training. In this work, we explore the feasibility of this approach before investing in computing for video pre-training. Our empirical findings across multiple video understanding tasks suggest that strong temporal performance can emerge without large-scale video pre-training, motivating future work on recurrent video foundation models obtained by pre-training a temporal module on top of a frozen image foundation model. Code: this https URL .
| Comments: | Accepted to CVPR 2026 Workshops CV4Smalls |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV) |
| Cite as: | arXiv:2605.19137 [cs.CV] |
| (or arXiv:2605.19137v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2605.19137 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Svetlana Orlova [view email]
[v1]
Mon, 18 May 2026 21:35:09 UTC (100 KB)
— Originally published at arxiv.org
More from arXiv cs.CV
See more →GeoSym127K: Scalable Symbolically-verifiable Synthesis for Multimodal Geometric Reasoning
GeoSym127K introduces a scalable neuro-symbolic framework for enhanced geometric reasoning in multimodal models.