SAM 2
PulseAugur coverage of SAM 2 — every cluster mentioning SAM 2 across labs, papers, and developer communities, ranked by signal.
4 day(s) with sentiment data
-
New VQS Benchmark and Method Introduced for Precise Video Object Segmentation
Researchers have introduced Visual Query Segmentation (VQS), a new method for precisely segmenting all instances of an object within an untrimmed video using a visual query. This approach offers more comprehensive and p…
-
UnSAMv2 enhances SAM with self-supervised learning for granular segmentation
Researchers have developed UnSAMv2, a novel self-supervised learning approach that significantly enhances the Segment Anything Model (SAM) family's ability to control segmentation granularity. Unlike previous methods th…
-
SAM3RL uses reinforcement learning to boost visual object tracking
Researchers have developed SAM3RL, a new extension for the Segment Anything Model 3 (SAM 3) that enhances visual object tracking. This model utilizes reinforcement learning to control its memory bank, deciding which fra…
-
New AI method estimates food volume from egocentric video without scale reference
Researchers have developed a novel method for estimating the volume of food from egocentric stereo video, eliminating the need for manual scale calibration or reference objects. This approach utilizes FoundationStereo f…
-
New method audits video model memory specificity
Researchers have developed a new method to audit the specificity of memory in video models, distinguishing between the general benefit of memory and the specific content retrieved. This technique, involving read-time me…
-
New SSS Method Enhances Medical Image Segmentation Using SAM-2
Researchers have developed SSS (Semi-Supervised SAM-2), a novel approach for semi-supervised medical image segmentation that leverages the feature extraction capabilities of SAM-2. This method aims to improve segmentati…
-
FoundYou unified model merges segmentation and retrieval tasks
Researchers have introduced FoundYou, a unified framework that combines personalized segmentation and retrieval tasks. Built upon the Segment Anything 2 (SAM 2) model, FoundYou leverages its ability to preserve object i…
-
FoundYou unified model integrates segmentation and retrieval tasks
Researchers have introduced FoundYou, a unified framework that integrates personalized segmentation and retrieval tasks by leveraging the instance-level cues captured by the Segment Anything 2 (SAM 2) model. This approa…
-
New SAM-H method sets state-of-the-art in planar object tracking
Researchers have developed SAM-H, a novel method for estimating homography poses for planar object tracking by leveraging segmentation mask contours. This approach, when combined with masks from SAM 2, achieves state-of…
-
New EditCLEVR benchmark tests AI object-centric representation faithfulness
Researchers have introduced EditCLEVR, a new benchmark designed to evaluate the compositional faithfulness of object-centric representations in AI models. This benchmark uses paired scenes with controlled semantic edits…
-
New Pipeline Enables Interactive 3D Asset Segmentation for Content Creation
Researchers have developed a human-in-the-loop pipeline for segmenting 3D assets into a 2D parameterized atlas. This method utilizes SAM 2 and Label Studio for interactive segmentation of rendered views, which are then …
-
SAMOSA framework adapts SAM 2 for advanced visual object tracking
Researchers have developed SAMOSA, a novel tracking framework that enhances the capabilities of the SAM 2 vision foundation model for complex visual object tracking. SAMOSA explicitly incorporates motion dynamics, geome…
-
TinySAM 2 offers efficient video segmentation with extreme memory compression
Researchers have developed TinySAM 2, a more efficient version of the Segment Anything Model 2 (SAM 2) for video segmentation and object tracking. TinySAM 2 employs a memory quality management mechanism and joint spatia…
-
New benchmarks evaluate geometric and physical consistency in video generation
Two new research papers introduce benchmarks for evaluating the geometric and physical consistency of generated videos. PDI-Bench focuses on assessing 3D structure and motion plausibility, while MechVerse specifically t…
-
SAM 3: The Eyes for AI — Nikhila & Pengchuan (Meta Superintelligence), ft. Joseph Nelson (Roboflow)
Meta AI has released SAM 3, a significant advancement in their Segment Anything project, capable of concept segmentation, detection, and tracking in images and video using natural language prompts. This new model achiev…