A developer says a computer-vision demo combining Meta's SAM 3.1 and DINOv3 kept track of two nearly identical surgical tools through partial occlusion and a long absence. In a video posted on X, the developer says the system kept tracking the tools as a forearm swept across the frame, then identified which tool returned after it had been gone for 15 seconds.
The models bring different documented capabilities to the setup. Meta describes SAM 3.1 as a model for detecting, segmenting and tracking objects in images and video. The company says DINOv3 produces dense visual features that support tasks including object detection, semantic segmentation and object tracking.
According to the developer, the demonstration ran on a single NVIDIA A100 through Hugging Face Jobs. Hugging Face's official account later reposted the clip, but the repost does not independently verify the result.
This is a personal demonstration, not a clinical study or product validation. It shows a promising result in one difficult tracking scene, but the supplied evidence does not establish reliability in real surgical settings.
