VOLEX-Fusion: Reliability-Aware Gated Fusion for Robust Event-Based Monocular Depth Estimation

Şebnem Sarıözkan, Alper Yegenoglu, Erdal Kayacan

Keywords: Event-Based Vision, Deep Learning for Visual Perception, Perception in Aerial Systems

TL;DR: VOLEX-Fusion is a monocular framework that adaptively fuses RGB and event data, prioritizing high-temporal event streams to maintain accurate depth estimation in low-light or high-motion blur scenarios.

Abstract: We introduce VOLEX-Fusion, a monocular depth estimation framework designed to maximize the utility of sparse event data through an adaptive gated fusion with RGB frames. The core innovation lies in the network’s ability to act as a reliability-aware arbitrator for depth inference; in scenarios where RGB quality degrades—due to severe motion blur or low-light—the fusion mechanism intelligently shifts its priority to the high-temporal event stream to maintain accurate depth consistency. By formulating depth estimation as a sensor-aware weighting problem, VOLEX-Fusion effectively bridges the gap between dense photometric textures and sparse event signals, ensuring robust topographic reconstruction even when the primary visual modality becomes unreliable.

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Submission Number: 6