aiqbench analysis
This research advances self-supervised learning by shifting focus from whole-image representations to per-object alignment using mask guidance. By extending the LeJEPA objective to object-level tasks without relying on negative samples, the method improves feature extraction efficiency for downstream applications like tracking and segmentation. This approach offers a more data-efficient pathway for training robust encoders in complex scenes where every image must yield maximum signal.
Mask-supervised Object-centric Representation Learning with LeJEPA
arXiv:2607.02404v2 Announce Type: replace-cross Abstract: Self-supervised image encoders deliver strong features for downstream tasks but need many images for training. A natural remedy to counter this is to make each image count for more. A scene contains many objects, and given masks from human annotators or an off-the-shelf segmentation model, pre-training can focus on aligning per-object rather than image-wide representations, extracting more signal from every image. Existing mask-supervised methods do this through reconstruction or contrastive losses that leverage negative objects. We instead use two separate projection spaces for the alignment. In a \emph{semantic space}, per-object representations from different views are aligned. To avoid collapse, instead of using negative objects, which requires category definitions, we extend the negative-free LeJEPA objective and show that its distributional anti-collapse regularizer ports naturally from whole images to the variable-sized set of objects in a scene. In an \emph{instance space}, a contrastive loss separates per-object representations from their context and co-occurring instances, including those of the same category. To separate object representations from their context, we copy objects and paste them into other contexts, where each pasted copy serves as an additional view of the original object. Trained on COCO with ground-truth masks, our method outperforms image-level and mask-guided baselines on tracking (DAVIS), classification (ImageNet-1k) and re-identification (NAVI), matches the best of them on semantic segmentation (ADE20k) and keeps its lead over image-level LeJEPA and a supervision-matched alternative on COCO fractions down to 256 images.
Abstract:Self-supervised image encoders deliver strong features for downstream tasks but need many images for training. A natural remedy to counter this is to make each image count for more. A scene contains many objects, and given masks from human annotators or an off-the-shelf segmentation model, pre-training can focus on aligning per-object rather than image-wide representations, extracting more signal from every image. Existing mask-supervised methods do this through reconstruction or contrastive losses that leverage negative objects. We instead use two separate projection spaces for the alignment. In a \emph{semantic space}, per-object representations from different views are aligned. To avoid collapse, instead of using negative objects, which requires category definitions, we extend the negative-free LeJEPA objective and show that its distributional anti-collapse regularizer ports naturally from whole images to the variable-sized set of objects in a scene. In an \emph{instance space}, a contrastive loss separates per-object representations from their context and co-occurring instances, including those of the same category. To separate object representations from their context, we copy objects and paste them into other contexts, where each pasted copy serves as an additional view of the original object. Trained on COCO with ground-truth masks, our method outperforms image-level and mask-guided baselines on tracking (DAVIS), classification (ImageNet-1k) and re-identification (NAVI), matches the best of them on semantic segmentation (ADE20k) and keeps its lead over image-level LeJEPA and a supervision-matched alternative on COCO fractions down to 256 images.
Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
arXiv:2607.02404 [cs.CV]
(or arXiv:2607.02404v2 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2607.02404 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 332 words
Related coverage
LeVJEPA: Efficient & Scalable Video Pretraining without the Heuristics
arXiv · Aug 28, 2026
Action-Slot: Structured Action-Centric Representation Learning for Multi-Agent Atomic Activity Understanding
arXiv · Sep 22, 2026
GORDON: Graph-based Object-centric Rewards for Decomposition of Long-Horizon Manipulation
arXiv · Aug 5, 2026
ParticleSplat: Self-supervised Object-centric Latent Particle Splatting
arXiv · Sep 18, 2026