Egocentric and Exocentric Methods: A Short Survey

2024-10-27

AI Summary

This work addresses the fundamental limitation of isolated modeling between egocentric and exocentric vision. We propose and systematically establish a novel ego-exo joint learning paradigm. Methodologically, we introduce cross-view alignment, feature transfer, synchronized dataset construction, and multimodal joint representation learning. Our study comprehensively surveys core challenges, benchmark datasets, task formulations, and state-of-the-art models in this emerging field for the first time. We empirically demonstrate strong complementarity between the two perspectives in embodied perception and formally characterize their distinct yet synergistic roles in visual understanding for AI agents. Furthermore, we construct the first structured knowledge graph and a reproducible experimental roadmap for ego-exo learning. These contributions provide both theoretical foundations and practical guidelines for multi-perspective video understanding, thereby advancing the development of next-generation embodied intelligent agents.

Technology Category

• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Machine Learning: Multimodal Learning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Egocentric vision captures the scene from the point of view of the camera wearer, while exocentric vision captures the overall scene context. Jointly modeling ego and exo views is crucial to developing next-generation AI agents. The community has regained interest in the field of egocentric vision. While the third-person view and first-person have been thoroughly investigated, very few works aim to study both synchronously. Exocentric videos contain many relevant signals that are transferrable to egocentric videos. This paper provides a timely overview of works combining egocentric and exocentric visions, a very new but promising research topic. We describe in detail the datasets and present a survey of the key applications of ego-exo joint learning, where we identify the most recent advances. With the presentation of the current status of the progress, we believe this short but timely survey will be valuable to the broad video-understanding community, particularly when multi-view modeling is critical.

Problem

• Jointly modeling ego and exo views for AI agents
• Studying egocentric and exocentric visions synchronously
• Surveying applications and datasets for ego-exo joint learning

Innovation

• Jointly modeling ego and exo views
• Transferring signals between exocentric and egocentric videos
• Surveying datasets for ego-exo joint learning