Every9D-21M supplies 21.8M real-world 9D pose annotations for 700 everyday categories by propagating manual canonical poses through cross-instance alignment in object-centric videos and verifying them multiview.
super hub Mixed citations
Title resolution pending
Mixed citation behavior. Most common role is background (43%).
597
Pith papers citing it
4
external citations · Pith
Background
43% of classified citations
abstract
Self-supervised learning holds the promise of eliminating the need for manual data annotation, enabling models to scale effortlessly to massive datasets and larger architectures. By not being tailored to specific tasks or domains, this training paradigm has the potential to learn visual representations from diverse sources, ranging from natural to aerial images -- using a single algorithm. This technical report introduces DINOv3, a major milestone toward realizing this vision by leveraging simple yet effective strategies. First, we leverage the benefit of scaling both dataset and model size by careful data preparation, design, and optimization. Second, we introduce a new method called Gram anchoring, which effectively addresses the known yet unsolved issue of dense feature maps degrading during long training schedules. Finally, we apply post-hoc strategies that further enhance our models' flexibility with respect to resolution, model size, and alignment with text. As a result, we present a versatile vision foundation model that outperforms the specialized state of the art across a broad range of settings, without fine-tuning. DINOv3 produces high-quality dense features that achieve outstanding performance on various vision tasks, significantly surpassing previous self- and weakly-supervised foundation models. We also share the DINOv3 suite of vision models, designed to advance the state of the art on a wide spectrum of tasks and data by providing scalable solutions for diverse resource constraints and deployment scenarios.
hub tools
citation-role summary
method 38
background 37
baseline 9
dataset 4
other 1
citation-polarity summary
claims ledger
- abstract Self-supervised learning holds the promise of eliminating the need for manual data annotation, enabling models to scale effortlessly to massive datasets and larger architectures. By not being tailored to specific tasks or domains, this training paradigm has the potential to learn visual representations from diverse sources, ranging from natural to aerial images -- using a single algorithm. This technical report introduces DINOv3, a major milestone toward realizing this vision by leveraging simple yet effective strategies. First, we leverage the benefit of scaling both dataset and model size by
authors
co-cited works
representative citing papers
LeJEPA achieves linear identifiability of latent variables uniquely when the latents are Gaussian in worlds with stationary additive-noise transitions.
Dimension d = O(m^{-2} log n) nearly achieves the optimal margin m^rd(+∞, A) for retrieval embeddings, with matching lower bounds showing d = O(k log(n/k)) suffices and is necessary for m = Θ(k^{-1/2}) on k-sparse query matrices.
RICA replaces ICA's global generative model with local Riemannian geometry, introducing a disentanglement tensor based on the Hessian of the log-likelihood and Ricci curvature to measure pointwise disentanglement, which recovers sources across manifolds in controlled tests.
CanViT is the first task- and policy-agnostic AVFM pretrained via passive-to-active dense latent distillation on 13.2M scenes and 1B random glimpses, achieving 38.5% ADE20K mIoU in one glimpse and 84.5% ImageNet-1k top-1 after fine-tuning.
ConceptPose delivers state-of-the-art zero-shot relative pose estimation by matching open-vocabulary 3D concept vectors derived from VLM saliency maps, beating the strongest baseline by 62% in ADD(-S) without training.
A 5B-parameter latent diffusion model generates real-time four-player Rocket League matches conditioned on all players' actions, staying stable far beyond its training horizon.
A training-free prototype memory-guided framework for multi-class prenatal ultrasound anomaly classification and localization using few reference images per class, validated on a 9-category multi-center dataset.
Introduces the CHC dataset of 3 m continuous canopy height differences with uncertainties and the uncertainty-aware change regression task for fine-tuning GFMs on PlanetScope time series.
Active Spatial Guidance replaces injected positional embeddings in ViTs with a training-only 2D coordinate regression loss on final-layer tokens, yielding better results than learned absolute or rotary embeddings on ImageNet-100, ADE20K, and Hypersim under matched training.
GEAR jointly trains VQ tokenizer and AR generator end-to-end via dual hard/soft read-out and representation alignment, achieving up to 10x faster ImageNet gFID convergence than LlamaGen-REPA while generalizing across quantizers and to text-to-image.
SemCityLoc achieves aerial 6DoF localization via semantic-geometric alignment of monocular depth and surfaces with LoD1-LoD3 city models, cutting mean positional error to 2.62m and boosting recall up to 36% on a new real-world benchmark.
OctoSense supplies a large multimodal robotics dataset and a late-fusion masked autoencoder that runs fast and outperforms image-only models on optical flow, depth, segmentation, and ego-motion tasks while remaining robust under sensor degradation.
Introduces SmellNet-V synthetic visuo-olfactory dataset and See & Sniff self-supervised framework that learns aligned representations and produces smell saliency maps.
Block-sparse featurizers recover visual concepts as two- to four-dimensional manifolds and describe activations more compactly than direction-based methods via minimum-description-length comparison.
MATCH is the first flow matching method for multi-view anomaly detection, reporting SOTA results on Real-IAD and the first comprehensive evaluation on MANTA-Tiny while enabling real-time use by omitting the divergence term.
REDI-Match uses rotation-equivariant distillation to transfer VFM semantics into a strictly equivariant encoder plus an entropy-driven alignment module, claiming SOTA accuracy and 1.9x speed on rotation-heavy benchmarks.
HeRA aligns least-aligned attention heads in MLLMs using an MKNN-based contrastive objective to preserve cross-modal topological structure, yielding gains on vision-centric tasks and reduced hallucinations across 18 benchmarks.
UniverSat is a ViT-style model with a universal patch encoder enabling self-supervised training on heterogeneous multimodal Earth observation data from varying resolutions and sensors.
SSL representations for echo LV segmentation and EF estimation recover substantially more task information with expressive downstream probes than with heuristic extraction, nearing supervised U-Net performance on EchoNet-Dynamic.
HiMatch-AD proposes DINOv3-driven hierarchical matching with uncertainty-based fusion for training-free medical anomaly detection and reports outperformance on the BMAD benchmark.
A method that treats 3D box pairs as exact transformation specs, adds a depth-aware floor reference, and trains an image generator on synthetic scenes plus Objectron videos to perform large 3D edits on real photographs.
FlowBender introduces closed-loop training that lets conditional flow models learn correction policies from their own task-specific alignment errors, outperforming supervised and guidance baselines on fidelity and plausibility.
EFIQA uses unsupervised masked anatomical inpainting to learn normal fundus structures and produces spatial quality maps via a shallow adapter on a frozen foundation model without any quality supervision.
citing papers explorer
-
Every9D-21M: Large-Scale Real-World 9D Canonicalization of Everyday Objects
Every9D-21M supplies 21.8M real-world 9D pose annotations for 700 everyday categories by propagating manual canonical poses through cross-instance alignment in object-centric videos and verifying them multiview.
-
When Does LeJEPA Learn a World Model?
LeJEPA achieves linear identifiability of latent variables uniquely when the latents are Gaussian in worlds with stationary additive-noise transitions.
-
Is Dimensionality a Barrier for Retrieval Models?
Dimension d = O(m^{-2} log n) nearly achieves the optimal margin m^rd(+∞, A) for retrieval embeddings, with matching lower bounds showing d = O(k log(n/k)) suffices and is necessary for m = Θ(k^{-1/2}) on k-sparse query matrices.
-
Disentanglement Beyond Generative Models with Riemannian ICA
RICA replaces ICA's global generative model with local Riemannian geometry, introducing a disentanglement tensor based on the Hessian of the log-likelihood and Ricci curvature to measure pointwise disentanglement, which recovers sources across manifolds in controlled tests.
-
CanViT: Toward Active-Vision Foundation Models
CanViT is the first task- and policy-agnostic AVFM pretrained via passive-to-active dense latent distillation on 13.2M scenes and 1B random glimpses, achieving 38.5% ADE20K mIoU in one glimpse and 84.5% ImageNet-1k top-1 after fine-tuning.
-
ConceptPose: Training-Free Zero-Shot Object Pose Estimation using Concept Vectors
ConceptPose delivers state-of-the-art zero-shot relative pose estimation by matching open-vocabulary 3D concept vectors derived from VLM saliency maps, beating the strongest baseline by 62% in ADD(-S) without training.
-
Multiplayer Interactive World Models with Representation Autoencoders
A 5B-parameter latent diffusion model generates real-time four-player Rocket League matches conditioned on all players' actions, staying stable far beyond its training horizon.
-
Prototype Memory-Guided Training-Free Anomaly Classification and Localization in Prenatal Ultrasound
A training-free prototype memory-guided framework for multi-class prenatal ultrasound anomaly classification and localization using few reference images per class, validated on a 9-category multi-center dataset.
-
Uncertainty-aware tree height change regression
Introduces the CHC dataset of 3 m continuous canopy height differences with uncertainties and the uncertainty-aware change regression task for fine-tuning GFMs on PlanetScope time series.
-
Active Spatial Guidance: Eliminating Injected Positional Mechanisms in Vision Transformers
Active Spatial Guidance replaces injected positional embeddings in ViTs with a training-only 2D coordinate regression loss on final-layer tokens, yielding better results than learned absolute or rotary embeddings on ImageNet-100, ADE20K, and Hypersim under matched training.
-
GEAR: Guided End-to-End AutoRegression for Image Synthesis
GEAR jointly trains VQ tokenizer and AR generator end-to-end via dual hard/soft read-out and representation alignment, achieving up to 10x faster ImageNet gFID convergence than LlamaGen-REPA while generalizing across quantizers and to text-to-image.
-
SemCityLoc: Aerial 6DoF Localization Using Semantic 3D City Models
SemCityLoc achieves aerial 6DoF localization via semantic-geometric alignment of monocular depth and surfaces with LoD1-LoD3 city models, cutting mean positional error to 2.62m and boosting recall up to 36% on a new real-world benchmark.
-
OctoSense: Self-Supervised Learning for Multimodal Robot Perception
OctoSense supplies a large multimodal robotics dataset and a late-fusion masked autoencoder that runs fast and outperforms image-only models on optical flow, depth, segmentation, and ego-motion tasks while remaining robust under sensor degradation.
-
See & Sniff: Learning Visuo-Olfactory Representations
Introduces SmellNet-V synthetic visuo-olfactory dataset and See & Sniff self-supervised framework that learns aligned representations and produces smell saliency maps.
-
Structuring Sparsity: Block-Sparse Featurizers Capture Visual Concept Manifolds
Block-sparse featurizers recover visual concepts as two- to four-dimensional manifolds and describe activations more compactly than direction-based methods via minimum-description-length comparison.
-
MATCH: Flow Matching for Multi-View Anomaly Detection
MATCH is the first flow matching method for multi-view anomaly detection, reporting SOTA results on Real-IAD and the first comprehensive evaluation on MANTA-Tiny while enabling real-time use by omitting the divergence term.
-
REDI-Match: Rotation-Equivariant Distillation for Efficient and Robust Dense Matching
REDI-Match uses rotation-equivariant distillation to transfer VFM semantics into a strictly equivariant encoder plus an entropy-driven alignment module, claiming SOTA accuracy and 1.9x speed on rotation-heavy benchmarks.
-
Mind the Heads: Topological Representation Alignment for Multimodal LLMs
HeRA aligns least-aligned attention heads in MLLMs using an MKNN-based contrastive objective to preserve cross-modal topological structure, yielding gains on vision-centric tasks and reduced hallucinations across 18 benchmarks.
-
UniverSat: Resolution- and Modality-Agnostic Transformers for Earth Observation
UniverSat is a ViT-style model with a universal patch encoder enabling self-supervised training on heterogeneous multimodal Earth observation data from varying resolutions and sensors.
-
Evaluating self-supervised echocardiographic representations across downstream extraction strategies for left-ventricular segmentation and ejection fraction estimation
SSL representations for echo LV segmentation and EF estimation recover substantially more task information with expressive downstream probes than with heuristic extraction, nearing supervised U-Net performance on EchoNet-Dynamic.
-
HiMatch-AD: DINOv3-driven Hierarchical Matching for Training-free Medical Anomaly Detection
HiMatch-AD proposes DINOv3-driven hierarchical matching with uncertainty-based fusion for training-free medical anomaly detection and reports outperformance on the BMAD benchmark.
-
Thinking in Boxes: 3D Editing in Real Images Made Easy
A method that treats 3D box pairs as exact transformation specs, adds a depth-aware floor reference, and trains an image generator on synthetic scenes plus Objectron videos to perform large 3D edits on real photographs.
-
FlowBender: Feedback-Aware Training for Self-Correcting Conditional Flows
FlowBender introduces closed-loop training that lets conditional flow models learn correction policies from their own task-specific alignment errors, outperforming supervised and guidance baselines on fidelity and plausibility.
-
EFIQA: Explainable Fundus Image Quality Assessment via Anatomical Priors
EFIQA uses unsupervised masked anatomical inpainting to learn normal fundus structures and produces spatial quality maps via a shallow adapter on a frozen foundation model without any quality supervision.
-
Algebraic Dead Directions in LayerNorm Transformers: A Forward-Pass-Only Diagnostic at LLM Scale
The normalized inverse-scale direction of LayerNorm's affine parameters is an exact algebraic kernel of the post-final-norm centred activation covariance for any input distribution in LayerNorm transformers.
-
Scaling Generative Foundation Models for Chest Radiography with Rectified Flow Transformers
A 1.3B-parameter rectified flow transformer is the first generative foundation model for chest radiograph synthesis at billion-parameter scale, producing images indistinguishable from real ones to experts.
-
The Reward Was in Your Data All Along: Correcting Flow Matching with Discriminator-Guided RL
DRL trains a discriminator on data versus base-model samples in pretrained representation space and uses its logit as reward in KL-regularized RL, cutting guidance-free FID from 9.38 to 2.62 on SiT and similar gains on other backbones.
-
Assessing Sample Quality in Conditional Generation under Compositional Shift
Introduces a per-sample trust score combining global realism and attribute-wise faithfulness, estimable from training data alone, for assessing conditional generations under compositional shift.
-
STREAM: Stochastic Riemannian Flow Matching with Anisotropic Decoder for Digital Histopathology Image Generation
STREAM applies stochastic Riemannian flow matching on VFM-derived unit hypersphere latents with a novel anisotropic decoder to achieve SOTA reconstruction and generation on breast and colorectal cancer histopathology datasets.
-
Diff-CA: Separating Common and Salient Factors with Diffusion Models
A diffusion-based contrastive analysis method that decomposes conditioning into common and salient factors with weak supervision and proves identifiability of the additive model.
-
Instant-Fold: In-Context Imitation Learning for Deformable Object Manipulation
Instant-Fold enables execution of multiple deformable object manipulation modes from a single demonstration via a flow-matching transformer policy that transfers zero-shot from simulation to real robots.
-
Benchmarking Visual State Tracking in Multimodal Video Understanding
VSTAT benchmark shows state-of-the-art MLLMs perform far below humans and only modestly above answer-prior baselines on visual state tracking, failing at visual perception despite correct textual reasoning.
-
Diffusing in the Right Space: A Systematic Study of Latent Diffusability
A large-scale empirical study across tokenizers and diffusion backbones identifies Velocity Irreducible Variance (VIV) as one of the most stable predictors of latent diffusion generation quality.
-
$A^2$: Smaller Self-Supervised ViTs Localize Better than Larger Ones
Smaller self-supervised ViTs localize objects better via attention than larger ViTs, enabling A² to decouple localization from feature extraction for competitive performance on distribution-shifted benchmarks.
-
HiTokSR: A Coarse-to-Fine Tokenizer with Hierarchical Codebooks for High-Fidelity Real-World Image Super-Resolution
HiTokSR uses a coarse-to-fine hierarchical tokenizer with frequency-aware sub-codebooks, vision foundation model priors, and index perturbation to achieve state-of-the-art perceptual quality and fidelity in real-world image super-resolution.
-
Chameleon: Style-Content Disentangled Framework for Cross-Domain Object Compositing
Chameleon proposes the first large-scale cross-domain compositing dataset and a disentangled encoder plus gated diffusion transformer that outperforms prior in-domain and cross-domain methods on plausibility and fidelity.
-
3DCodeBench: Benchmarking Agentic Procedural 3D Modeling Via Code
3DCodeBench is a new benchmark evaluating 12 VLMs on translating multimodal prompts into procedural 3D modeling code, paired with 3DCodeArena for human preference rankings.
-
Cohort-Scale Neural Atlases of Ultrasound Video
A single canonical neural atlas is learned jointly over thousands of ultrasound frames from five cardiac and musculoskeletal datasets via DINOv3 features and per-video generative latent optimization embeddings to support annotation transfer.
-
Images as Tables: In-Context Learning with TabPFN for Low-Data Detection of AI-Generated Images
DINOv3-PCA-TabPFN outperforms prior detectors like LATTE in low-data and cross-generator transfer settings for AI image detection.
-
Learning Global Motion with Compact Gaussians for Feed-Forward 4D Reconstruction
C4G introduces compact timestamp-conditioned Gaussian query tokens that aggregate full temporal context to decode 3D Gaussians with timestamp-modulated positions for feed-forward 4D reconstruction from monocular video, plus a diffusion-based rendering module and extension to 4D feature fields.
-
Ego-METAS: Egocentric online Multimodal Energy-efficient Temporal Action Segmentation benchmark
Ego-METAS is a new benchmark providing unified egocentric video data, splits, features and baselines for online multimodal temporal action segmentation under hardware-representative energy constraints.
-
Structure over Pixels: Learning Variable-Length Visual Programs
STROP learns variable-length discrete visual programs for images by training a length head against frozen DINOv3 features in a four-phase curriculum while bypassing pixel reconstruction.
-
OmniRetriever: Any-to-Any Audio-Video-Text Retrieval via Fusion-as-Teacher Distillation
OmniRetriever-7B uses fusion-as-teacher distillation plus Tuple-InfoNCE to improve any-to-any audio-video-text retrieval over prior open and closed models.
-
EchoPilot: Training-Free Ultrasound Video Segmentation via Scale-Space Semantic Prompting and Reliability-Gated Memory
EchoPilot delivers state-of-the-art training-free ultrasound video segmentation from a single point prompt by introducing scale-space semantic prompting via S.E.E.D. and reliability-gated memory updates.
-
Benchmarking Pathology Foundation Models for Spatial Domain Understanding
SpaPath-Bench evaluates spatial representation in 19 pathology foundation models via spatial domain identification on 42 paired WSI-ST slides using three agreement criteria across 83K runs.
-
Uncertainty-DTW for Sequences and Visual Tokens
Uncertainty-DTW models pairwise correspondences with Normal distributions and uses an MLE objective with precision-weighted matching plus log-variance regularization for robust alignment of sequences and visual tokens.
-
Rethinking Continual Anomaly Detection on the Edge: Benchmarking Under Realistic Industrial Conditions
Introduces a unified benchmark for continual anomaly detection with discrete and continuous protocols plus a training-free DINOSaur method that outperforms prior CAD approaches with zero forgetting and sub-100ms edge inference.
-
GenRecon: Bridging Generative Priors for Multi-View 3D Scene Reconstruction
GenRecon lifts object-level generative priors to scene-scale reconstruction by chunking scenes and using projection-based conditioning on multi-view features, claiming 16% better results than prior methods.
-
DecQ: Detail-Condensing Queries for Enhanced Reconstruction and Generation in Representation Autoencoders
DecQ uses detail-condensing queries on shallow and deep VFM features to improve both reconstruction PSNR and generative convergence/FID in RAEs without fine-tuning the encoder.
-
H-Flow: Self-supervised Human Scene Flow via Physics-inspired Joint Multi-modal Learning
H-Flow learns dense human scene flow from monocular video via joint pose and depth prediction in a multi-head transformer, using physics-inspired geometric and biomechanical priors for self-supervision, and introduces the DynAct4D synthetic benchmark.