Roman Rädle

SAM 3 is not a referring expression segmentation model - this is by design. SAM 3 solves Promptable Concept Segmentation (PCS): segmenting objects using simple noun phrases (general categories + basic attributes) and optional exemplars. It's also a robust, composable primitive that works seamlessly with MLLMs. For example, SAM 3 + Gemini 2.5 Pro achieves zero-shot SOTA on RefCOCO+/RefCOCOg referring expression benchmarks. Wondering if this wasn't sufficiently clear from the paper (our bad if so). But also wondering why only referring expression is benchmarked and not any of the benchmarks we reported on—the ones SAM 3 is actually designed for?

i trained this computer vision model just by asking for "tennis players", using roboflow rapid and SAM3 the model picks out Alcaraz and Sinner, while correctly avoiding the ball boys and fans then I exported the model to python and wrote a script to filter the data, track movement history, and annotate the final video this was a quick experiment but I could expand the project to track player speeds, distance covered, shot count, etc. could use this for a sports analytics app that tracks live matches and calculates custom metrics

SOTA referral segmentation ✅

Encouraging to see medical applications of SAM 3 in just 1 week. We didn't prioritize medical use cases, so works like the MedSAM series that add real clinical data and knowledge are especially valuable. Medicine & science uses for SAM remain some of the most rewarding to see 🩺

MedSAM3: Segment Anything with Medical Concepts A new framework for text-promptable medical segmentation (PCS). It adapts SAM3 with medical concepts for diverse modalities (X-ray, MRI, CT, video), outperforming existing models & simplifying anatomical targeting.