DINOv3
PulseAugur coverage of DINOv3 — every cluster mentioning DINOv3 across labs, papers, and developer communities, ranked by signal.
- used by DagsHub 90%
- used by alphaXiv 90%
- used by CatalyzeX 90%
- used by ConvNeXt 90%
- used by ViT-L/16 90%
- used by ICRA 2026 GOOSE 2D Fine-Grained Semantic Segmentation Challenge 90%
- used by YOLOv12 80%
- used by Gotit.pub 70%
- used by ScienceCast 70%
- used by Sam3 70%
- competes with Sam3 70%
- used by vision transformer 70%
22 day(s) with sentiment data
-
LoRCA framework enables histology to HiP-CT image translation
Researchers have developed LoRCA (LoRA Cycle Adaptation), a novel framework for translating histology images to Hierarchical Phase-Contrast Tomography (HiP-CT) volumes. This method utilizes a frozen DINOv3 backbone with…
-
DINOv3 fine-tuned for medical image classification achieves SOTA results
Researchers have developed an efficient fine-tuning method for the DINOv3-H+ vision transformer, originally trained on natural images, to classify atypical mitotic figures (AMFs). By using low-rank adaptation (LoRA) and…
-
TriView-YOLO model enhances cavity detection in challenging soil conditions
Researchers have developed TriView-YOLO, a novel deep learning model designed for detecting subsurface cavities in challenging soft, high-water-content soils. This model utilizes a multi-view fusion approach, integratin…
-
New SRE-FER framework enhances facial expression recognition accuracy
Researchers have developed SRE-FER, a new framework designed to improve fine-grained facial expression recognition by addressing the issue of local evidence dilution. This problem occurs when global aggregation in found…
-
New DINO-3DRA framework improves 3D cerebral aneurysm segmentation
Researchers have developed DINO-3DRA, a novel framework for segmenting cerebral aneurysms in 3D rotational angiography (3DRA) data. The system effectively transfers semantic knowledge from 2D foundation models, like DIN…
-
Modern backbones boost AI for mammography classification and lesion localization
Researchers have explored the use of modern neural network backbones within a multi-task DETR framework to enhance mammography classification and lesion localization. The study found that advanced backbones like ConvNeX…
-
New AI framework enables autonomous robotic laparoscope control
Researchers have developed SurgLAT, a novel framework for controlling robotic laparoscopes autonomously. This system models the surgeon's evolving attention state in real-time, using a DINOv3 encoder and a causal latent…
-
Multimodal AI Classifies Red Deer Sex and Age from Drone Video
Researchers have developed a multimodal approach to classify sex and life stage in red deer using aerial RGB-thermal video. By fusing data from both sensors at every stage, the pipeline improves accuracy compared to usi…
-
New ECAD framework expands object detection beyond discrete instances
Researchers have introduced Expanded Class-Agnostic Detection (ECAD), a new framework designed to go beyond traditional object detection by identifying category-agnostic visual candidates that are not limited to discret…
-
SEED system offers explainable detection for AI-generated text forgeries
Researchers have developed SEED, a system designed to detect and explain AI-generated text forgeries. This system, which ranked third in the GenText-Forensics Challenge at ACM MM 2026, utilizes a Vision Transformer (ViT…
-
New methods improve reliability of AI in medical image segmentation
Researchers have developed methods to improve the reliability of in-context learning for medical image segmentation. They found that selecting support set exemplars based on visual similarity to the query image, rather …
-
New SO-OPF method precisely analyzes vision encoder changes
Researchers have developed a new method called Support Operation Factorization (SO-OPF) to analyze frozen vision encoders, aiming to precisely identify what changes and where within the encoder's operations. This techni…
-
Tarot-SAM3 framework enhances SAM3 for any referring expression segmentation
Researchers have developed Tarot-SAM3, a new framework designed to improve referring expression segmentation (RES) by enabling the Segment Anything Model 3 (SAM3) to handle any natural language expression. The framework…
-
New diffusion models enhance pathological image resolution for better diagnostics · 2 sources tracked
Two new research papers propose advanced diffusion models for enhancing the resolution of pathological images, aiming to improve diagnostic accuracy. S$^3$-Diff utilizes a Structural Semantic Synergy approach with speci…
-
New Recurrent Contrastive Learning method boosts imbalanced medical image classification
Researchers have introduced Recurrent Contrastive Learning (RCL), a novel method designed to improve imbalanced medical image classification. RCL aims to expand the feature representation of underrepresented classes by …
-
New frameworks boost open-vocabulary segmentation for remote sensing
Researchers have developed two new frameworks for open-vocabulary semantic segmentation in remote sensing. The first, DinoSplat-OV, adapts the DINOv3 model to this domain without fine-tuning, using modules for text-awar…
-
AI model LeDXA extracts disease risk and biological age from X-ray scans
Researchers have developed LeDXA, a self-supervised learning model that extracts health insights from dual-energy X-ray absorptiometry (DXA) scans. Trained on unlabeled DXA images, LeDXA predicts disease risk, biologica…
-
Foundation model pretraining strategies impact retinal imaging transferability
A new arXiv paper explores how different pretraining strategies for foundation models impact their effectiveness when transferred to ultra-widefield retinal imaging tasks. Researchers compared Vision Transformer encoder…
-
BladeYOLO framework enhances wind turbine defect detection with limited data
Researchers have developed BladeYOLO, a new framework designed to improve the detection of defects on wind turbine blades, particularly in scenarios with limited annotated data. The system integrates a Vision Transforme…
-
New World-Action Models Enhance Robot Manipulation and Generalization
Researchers have developed several new world-action models (WAMs) for robotic manipulation that aim to improve efficiency and robustness. LiLa-WAM focuses on a lightweight latent reasoning space for end-to-end training …