Scaling has not yet been convincingly demonstrated for pure self-supervised learning from video. However, prior work has focused evaluations on semantic-related tasks $\unicode{x2013}$ action classification, ImageNet classification, etc. In this paper we focus on evaluating self-supervised learning on non-semantic vision tasks that are more spatial (3D) and temporal (+1D = 4D), such as camera pose estimation, point and object tracking, and depth estimation. We show that by learning from very large video datasets, masked auto-encoding (MAE) with transformer video models actually scales, consistently improving performance on these 4D tasks, as model size increases from 20M all the way to the largest by far reported self-supervised video model $\unicode{x2013}$ 22B parameters. Rigorous apples-to-apples comparison with many recent image and video models demonstrates the benefits of scaling 4D representations.

We introduce a hierarchical probabilistic approach to go from a 2D image to multiview 3D: a diffusion "prior" models the unseen 3D geometry, which then conditions a diffusion "decoder" to generate novel views of the subject. We use a pointmap-based geometric representation in a multiview image format to coordinate the generation of multiple target views simultaneously. We facilitate correspondence between views by assuming fixed target camera poses relative to the source camera, and constructing a predictable distribution of geometric features per target. Our modular, geometry-driven approach to novel-view synthesis (called "unPIC") beats SoTA baselines such as CAT3D and One-2-3-45 on held-out objects from ObjaverseXL, as well as real-world objects ranging from Google Scanned Objects, Amazon Berkeley Objects, to the Digital Twin Catalog.

Current vision models typically maintain a fixed correspondence between their representation structure and image space. Each layer comprises a set of tokens arranged "on-the-grid," which biases patches or tokens to encode information at a specific spatio(-temporal) location. In this work we present Moving Off-the-Grid (MooG), a self-supervised video representation model that offers an alternative approach, allowing tokens to move "off-the-grid" to better enable them to represent scene elements consistently, even as they move across the image plane through time. By using a combination of cross-attention and positional embeddings we disentangle the representation structure and image structure. We find that a simple self-supervised objective--next frame prediction--trained on video data, results in a set of latent tokens which bind to specific scene structures and track them as they move. We demonstrate the usefulness of MooG's learned representation both qualitatively and quantitatively by training readouts on top of the learned representation on a variety of downstream tasks. We show that MooG can provide a strong foundation for different vision tasks when compared to "on-the-grid" baselines.

PaliGemma is an open Vision-Language Model (VLM) that is based on the SigLIP-So400m vision encoder and the Gemma-2B language model. It is trained to be a versatile and broadly knowledgeable base model that is effective to transfer. It achieves strong performance on a wide variety of open-world tasks. We evaluate PaliGemma on almost 40 diverse tasks including standard VLM benchmarks, but also more specialized tasks such as remote-sensing and segmentation.

We address the problem of multi-object 3D pose control in image diffusion models. Instead of conditioning on a sequence of text tokens, we propose to use a set of per-object representations, Neural Assets, to control the 3D pose of individual objects in a scene. Neural Assets are obtained by pooling visual representations of objects from a reference image, such as a frame in a video, and are trained to reconstruct the respective objects in a different image, e.g., a later frame in the video. Importantly, we encode object visuals from the reference image while conditioning on object poses from the target frame. This enables learning disentangled appearance and pose features. Combining visual and 3D pose representations in a sequence-of-tokens format allows us to keep the text-to-image architecture of existing models, with Neural Assets in place of text tokens. By fine-tuning a pre-trained text-to-image diffusion model with this information, our approach enables fine-grained 3D pose and placement control of individual objects in a scene. We further demonstrate that Neural Assets can be transferred and recomposed across different scenes. Our model achieves state-of-the-art multi-object editing results on both synthetic 3D scene datasets, as well as two real-world video datasets (Objectron, Waymo Open).

Learning structured representations of visual scenes is currently a major bottleneck to bridging perception with reasoning. While there has been exciting progress with slot-based models, which learn to segment scenes into sets of objects, learning configurational properties of entire groups of objects is still under-explored. To address this problem, we introduce Constellation, a network that learns relational abstractions of static visual scenes, and generalises these abstractions over sensory particularities, thus offering a potential basis for abstract relational reasoning. We further show that this basis, along with language association, provides a means to imagine sensory content in new ways. This work is a first step in the explicit representation of visual relationships and using them for complex cognitive procedures.

We present a slot-wise, object-based transition model that decomposes a scene into objects, aligns them (with respect to a slot-wise object memory) to maintain a consistent order across time, and predicts how those objects evolve over successive frames. The model is trained end-to-end without supervision using losses at the level of the object-structured representation rather than pixels. Thanks to its alignment module, the model deals properly with two issues that are not handled satisfactorily by other transition models, namely object persistence and object identity. We show that the combination of an object-level loss and correct object alignment over time enables the model to outperform a state-of-the-art baseline, and allows it to deal well with object occlusion and re-appearance in partially observable environments.

Recently developed deep learning models are able to learn to segment scenes into component objects without supervision. This opens many new and exciting avenues of research, allowing agents to take objects (or entities) as inputs, rather that pixels. Unfortunately, while these models provide excellent segmentation of a single frame, they do not keep track of how objects segmented at one time-step correspond (or align) to those at a later time-step. The alignment (or correspondence) problem has impeded progress towards using object representations in downstream tasks. In this paper we take steps towards solving the alignment problem, presenting the AlignNet, an unsupervised alignment module.

The ability to decompose scenes in terms of abstract building blocks is crucial for general intelligence. Where those basic building blocks share meaningful properties, interactions and other regularities across scenes, such decompositions can simplify reasoning and facilitate imagination of novel scenarios. In particular, representing perceptual observations in terms of entities should improve data efficiency and transfer performance on a wide range of tasks. Thus we need models capable of discovering useful decompositions of scenes by identifying units with such regularities and representing them in a common format. To address this problem, we have developed the Multi-Object Network (MONet). In this model, a VAE is trained end-to-end together with a recurrent attention network -- in a purely unsupervised manner -- to provide attention masks around, and reconstructions of, regions of images. We show that this model is capable of learning to decompose and represent challenging 3D scenes into semantically meaningful components, such as objects and background elements.

The field of reinforcement learning (RL) is facing increasingly challenging domains with combinatorial complexity. For an RL agent to address these challenges, it is essential that it can plan effectively. Prior work has typically utilized an explicit model of the environment, combined with a specific planning algorithm (such as tree search). More recently, a new family of methods have been proposed that learn how to plan, by providing the structure for planning via an inductive bias in the function approximator (such as a tree structured neural network), trained end-to-end by a model-free RL algorithm. In this paper, we go even further, and demonstrate empirically that an entirely model-free approach, without special structure beyond standard neural network components such as convolutional networks and LSTMs, can learn to exhibit many of the characteristics typically associated with a model-based planner. We measure our agent's effectiveness at planning in terms of its ability to generalize across a combinatorial and irreversible state space, its data efficiency, and its ability to utilize additional thinking time. We find that our agent has many of the characteristics that one might expect to find in a planning algorithm. Furthermore, it exceeds the state-of-the-art in challenging combinatorial domains such as Sokoban and outperforms other model-free approaches that utilize strong inductive biases toward planning.