A single DINOv3-pretrained ConvNeXt trained with a tuned recipe achieves 88.19 mAP on VeRi-Wild Small and 77.47 mAP on VeRi-Wild Large for vehicle re-identification, matching the strongest multi-branc…
The Department of Energy's Genesis Mission is using Meta's SAM 3 and DINOv3 vision models to turn raw X-ray scans into labeled 3D images in about 15 minutes, according to a July 21st Meta blog post. B…
Researchers show that the frozen predictor from Joint-Embedding Predictive Architectures (JEPAs) can be transferred across encoder families to improve occluded feature completion. In experiments, pair…
Researchers propose C-Norm (Cell-Distribution Normalization), a method that decouples abnormal and normal cells from ThinPrep Cytologic Test (TCT) images and re-synthesizes them to ensure uniform cell…
Researchers present AffectFlow-DINO, a multi-task learning system for the 11th ABAW challenge that uses a conditional rectified-flow head to model ambiguity in facial behavior, achieving a multi-task …
RF-DETR is the strongest starting point for most computer vision projects in 2026, topping both COCO and the real-world RF100-VL benchmark for object detection and segmentation, with SAM 3, DINOv3, GL…
A study testing 15 foundation models for mammography across 12 out-of-distribution datasets found that models like Mammo-FM and MaMA showed promise but struggled with generalization, while DINOv3, a v…
A new study from arXiv provides exact solutions for how linear representations emerge in neural networks during training, a process called abstraction. The researchers found that data and target geome…
Ant Group's embodied-AI unit Robbyant open-sourced LingBot-Vision, a family of self-supervised Vision Transformers designed for dense spatial perception, on Hugging Face under Apache-2.0. The 1.1B-par…
Researchers from General Intuition, Kyutai, and Epic Games released MIRA, a 5-billion-parameter latent diffusion world model that generates real-time 2v2 Rocket League matches at 20 FPS on a single GP…
Ant Group's embodied AI unit, Robbyant, launched LingBot-Depth 2.0 and LingBot-Vision, claiming the models enable robots to accurately perceive transparent objects like glass and mirrors. The company …
Researchers introduced block-sparse featurizers (BSFs) to capture visual concept manifolds in neural network representations, showing that concepts are low-dimensional geometric structures rather than…
Researchers introduced MR-DiffuSR, a 3D latent diffusion framework that uses high-resolution T1w structural priors to guide super-resolution of thick-slice FLAIR MRI scans, preventing anatomical hallu…
Researchers achieved first place in the ICRA 2026 GOOSE 2D Fine-Grained Semantic Segmentation Challenge by combining a DINOv3 ViT-L/16 backbone with a Mask2Former decoder and inference-time aggregatio…
Researchers introduced the Overhead Wildlife Locator (OWL), a weakly supervised density-estimation framework for aerial wildlife surveys that uses point-level labels instead of costly bounding-box ann…
Researchers have developed STROP, a discrete visual tokenizer that learns to generate variable-length structural scene representations, determining the optimal sequence length for each image in a sing…
Researchers have developed representation-conditioned diffusion models that generate synthetic training images by conditioning on learned representations from DINOv2, DINOv3, and CLIP, achieving a 10.…