A two-stage, training-free solution using Gemini-3.1 Pro and SAM3 ranked third in the MeViS-Text track of the 8th LSVOS Challenge, achieving J&F, J, F, N-acc., T-acc., and Final scores of 0.761, 0.736…
VOS-Agent, a collaborative framework built on SAM3, won first place in the MOSEv2 Track of the 8th LSVOS Challenge at ECCV 2026, achieving 69.82% on the official J&F metric. The framework uses a Targe…
A new method called Object-Part Hierarchical Reflective Grounding (OP-HRG) improves part-level visual grounding in multimodal large language models by introducing a coarse-to-fine reasoning pipeline t…
A new study from arXiv introduces GenCeption, a model that leverages a pre-trained video generative diffusion backbone to perform general-purpose vision tasks, achieving state-of-the-art results in de…
Meta's SAM3 (Segment Anything Model 3) enables open-vocabulary object detection from text prompts without labeled data or training. Roboflow Workflows integrates SAM3 to build a warehouse asset search…
A user resolved a deployment issue with the SAM3 model in Microsoft Azure after identifying an internal corporate policy restriction that prevented the deployment from reflecting correctly. The policy…
Researchers propose Dual-Adaptive SAM3 (DA-SAM3), a framework for parameter-efficient medical image segmentation that uses a Dynamic Expert Router and Decomposed Parameterized Experts to reduce MoE ov…
Roboflow released a text prompt object detection workflow using SAM3 that lets users find objects by typing descriptions instead of training models, enabling rapid prototyping for plant disease detect…