AI Guide Dog: Egocentric Path Prediction on Smartphone

2025-01-14

AI Summary

To address the indoor and outdoor navigation needs of visually impaired individuals, this paper proposes the first lightweight, first-person visual navigation system that unifies goal-directed navigation and target-free exploration. Methodologically: (1) a pure-vision multi-label classification model is designed to generate high-level semantic navigation commands in real time; (2) GPS localization is innovatively fused with semantic instructions for robust outdoor navigation; (3) an uncertainty-aware indoor path prediction mechanism is introduced to handle ambiguous, multi-path scenarios; and (4) the system is deployed on mobile devices using a custom-built, multi-scenario navigation dataset. Experiments demonstrate real-time inference on smartphones, achieving new state-of-the-art performance for blind navigation. The system is validated across diverse indoor and outdoor environments for safety, generalizability, and practical utility. All data, evaluation protocols, and deployment artifacts are publicly released.

Technology Category

• Computer Vision: Multi-modal Vision
• Intelligent Robots: Localization, Mapping, and Navigation
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems

Abstract

This paper introduces AI Guide Dog (AIGD), a lightweight egocentric navigation assistance system for visually impaired individuals, designed for real-time deployment on smartphones. AIGD addresses key challenges in blind navigation by employing a vision-only, multi-label classification approach to predict directional commands, ensuring safe traversal across diverse environments. We propose a novel technique to enable goal-based outdoor navigation by integrating GPS signals and high-level directions, while also addressing uncertain multi-path predictions for destination-free indoor navigation. Our generalized model is the first navigation assistance system to handle both goal-oriented and exploratory navigation scenarios across indoor and outdoor settings, establishing a new state-of-the-art in blind navigation. We present methods, datasets, evaluations, and deployment insights to encourage further innovations in assistive navigation systems.

Problem

• Visual Impairment
• Indoor and Outdoor Navigation
• Smart Navigation System

Innovation

• AI Guide Dog
• Indoor Navigation
• Image Recognition