Segment Anything Model 3
PulseAugur coverage of Segment Anything Model 3 — every cluster mentioning Segment Anything Model 3 across labs, papers, and developer communities, ranked by signal.
- 2026-05-22 product_launch Meta AI launched Segment Anything Model 3 (SAM 3), an advanced model for object detection, segmentation, and tracking. source
4 day(s) with sentiment data
-
Dyna3 framework enables training-free 4D dynamic scene reconstruction
Researchers have developed Dyna3, a novel framework that enables dynamic 4D scene reconstruction using depth foundation models without requiring any fine-tuning. This method leverages the implicit motion-discriminative …
-
Feyn releases MultiMatte, a promptable image background removal model
Feyn has released MultiMatte, a new image background removal model that uses natural language prompts to identify and isolate objects. Built upon Meta's Segment Anything Model 3 (SAM 3), MultiMatte employs low-rank fine…
-
LeCor method improves 3D lung tumor segmentation with meta-learned training
Researchers have developed LeCor, a novel method for improving 3D lung tumor segmentation in computed tomography (CT) scans. LeCor utilizes meta-learned test-time training, where each clinician's correction acts as a tr…
-
New SAM-Radar framework enhances object tracking with multimodal sensor fusion
Researchers have introduced RGBTR-Motion, a new benchmark dataset for moving-object segmentation and tracking that integrates RGB, thermal, and radar streams. They also developed SAM-Radar, a framework that leverages th…
-
New CLON method enhances zero-shot 6D pose estimation
Researchers have developed CLON (Cue-Calibrated Linguistic Object Onboarding), a novel front-end for zero-shot 6D pose estimation. CLON constructs a linguistic semantic memory to guide proposal generation and uses objec…
-
ENEAS method enhances video segmentation and instance tracking
Researchers have introduced ENEAS, a novel method designed to improve instance tracking and semantic segmentation in videos and other data. ENEAS addresses limitations in current text-promptable segmentation models, suc…
-
ENEAS method enhances text-prompted instance tracking and semantic discovery
Researchers have introduced ENEAS, a novel method designed to improve text-prompted instance tracking and open-concept semantic discovery in segmentation models. ENEAS addresses limitations found in current foundation m…
-
New MariSat dataset targets maritime object segmentation in satellite imagery
Researchers have introduced MariSat, a new dataset designed for instance segmentation of maritime objects in satellite and aerial imagery. The dataset comprises 1260 images annotated at the pixel level for eight distinc…
-
New MemMTL framework enhances multi-task dense prediction with prototype memory
Researchers have developed MemMTL, a new framework for multi-task dense prediction that utilizes a learnable task-state prototype memory. This memory refines a compact task state derived from global visual context, whic…
-
New method enhances zero-shot segmentation for UAV tower inspections
Researchers have developed a new method called Saliency-Depth Conditioning to improve zero-shot segmentation of communication-tower components in cluttered UAV imagery. This approach combines visual saliency with monocu…
-
New method uses SAM 3 for surgical landmark localization
Researchers have developed a new method for localizing functional landmarks in surgical videos by leveraging the Segment Anything Model 3 (SAM 3). This approach uses SAM 3's structural prior to provide dense instrument-…
-
SAM3Dual enhances SAM 3 for video object segmentation without fine-tuning
Researchers have developed SAM3Dual, a novel approach that enhances the Segment Anything Model 3 (SAM 3) for video object segmentation. This method, which achieved third place in the MOSEv2 track at the 8th Large-scale …
-
New framework uses visual memory to guide AI models in imaging tasks
Researchers have introduced Retrieval-Augmented Visual Prompting (RAVP), a novel framework designed to guide foundation models in specialized tasks like two-photon calcium imaging. Instead of fine-tuning model weights, …
-
New SAM3-based method enables annotation-free surgical instrument segmentation
Researchers have developed a novel two-stage framework for segmenting surgical instruments in endoscopic images without requiring manual annotation. This approach utilizes the Segment Anything Model 3 (SAM3) with a gene…
-
New framework improves spacecraft segmentation using foundation models
Researchers have developed GeoDistill-Refine, a novel two-stage framework designed to improve the accuracy of spacecraft segmentation using foundation models. This method addresses geometric errors in pseudo-masks gener…
-
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 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…
-
SAM3 adapted for surgical segmentation using parameter-efficient LoRA
Researchers have developed a parameter-efficient adaptation of the Segment Anything Model 3 (SAM3) specifically for surgical concept segmentation. This new method, utilizing Low-Rank Adaptation (LoRA), significantly red…
-
New AI workflow maps farmland extent using satellite imagery and SAM 3
Researchers have developed a new workflow to map farmland extent and boundaries using 1-meter NAIP imagery. The method combines a Residual U-Net model, trained with a Dice-dominant loss, and a Segment Anything Model (SA…
-
New method enhances open-vocabulary segmentation for remote sensing
Researchers have developed Prompt-Calibrated SAM 3 (ProC-SAM3), a novel approach to open-vocabulary semantic segmentation in remote sensing. This method addresses limitations in existing SAM 3-based techniques by creati…