Augmented Object Intelligence with XR-Objects

TL;DR

This paper introduces Augmented Object Intelligence (AOI) to embed digital capabilities into physical objects within XR, demonstrated by XR-Objects, an open-source prototype. The system combines real-time object segmentation, 3D localization, and a per-object Multimodal Large Language Model to deliver context-aware actions anchored to objects without pre-registration. A smartphone-based implementation in Unity/AR Foundation is evaluated against a state-of-the-art AI assistant, showing faster task completion times and positive usability feedback across cooking, shopping, discovery, productivity, learning, and IoT scenarios. The work underlines the potential of OAIs to transform everyday interactions by making objects themselves intelligent interfaces and outlines future directions for head-mounted deployments, hallucination mitigation, and AGI-driven interface generation.

Abstract

Seamless integration of physical objects as interactive digital entities remains a challenge for spatial computing. This paper explores Augmented Object Intelligence (AOI) in the context of XR, an interaction paradigm that aims to blur the lines between digital and physical by equipping real-world objects with the ability to interact as if they were digital, where every object has the potential to serve as a portal to digital functionalities. Our approach utilizes real-time object segmentation and classification, combined with the power of Multimodal Large Language Models (MLLMs), to facilitate these interactions without the need for object pre-registration. We implement the AOI concept in the form of XR-Objects, an open-source prototype system that provides a platform for users to engage with their physical environment in contextually relevant ways using object-based context menus. This system enables analog objects to not only convey information but also to initiate digital actions, such as querying for details or executing tasks. Our contributions are threefold: (1) we define the AOI concept and detail its advantages over traditional AI assistants, (2) detail the XR-Objects system's open-source design and implementation, and (3) show its versatility through various use cases and a user study.

Figures (14)

• Figure 1: Example of an interaction with a conventional multimodal AI assistant. While the model clearly has the capacity to produce reasonable scene understanding when an image and a prompt is provided as input, it fails in providing an anchored output that ties back to the original multimodal prompt.
• Figure 2: The landscape of physical object interactions in XR classified across two dimensions: anchoring and content.
• Figure 3: The XR-Objects processing pipeline combines MediaPipe and ARCore for object detection and spatial tracking, respectively, integrates an MLLM for object-specific metadata retrieval and interaction, and renders UI content in 3D space.
• Figure 4: XR-Objects instantiates a dedicated MLLM instance for identified object in the scene. Object comparisons are executed by stitching together the relevant objects in the scene before passing the query to an MLLM instance.
• Figure 5: User study setup with mock grocery store (a) and at-home (b) environments. Examples of using XR-Objects in each case are shown in (c) and (d), respectively.