apply for access
I am applying for access to SAM 3 for academic research and prototype development in AR-assisted human-object interaction.
My project aims to build an AR-style application that takes natural-language instructions from users, identifies the relevant object in the scene, generates an accurate segmentation mask, and highlights the target object in real time. The system will also provide concise textual guidance, such as indicating which object should be moved, picked up, or interacted with.
SAM 3 will be used as the core visual segmentation module for promptable object localization and object highlighting in indoor manipulation scenarios. I plan to evaluate its performance on common objects, cluttered scenes, and language-guided interaction tasks, and compare it with existing segmentation models such as SAM 2.
The model checkpoint will be used locally for non-commercial research experiments and prototype evaluation only.