M^2C-EvDet: Multi-Domain Multi-Order Cross-Modal Knowledge Distillation for Event-based Object Detection

2026-06-23

AI Summary

This work addresses the significant performance gap in object detection between event-based and frame-based methods, primarily caused by the sparsity of event data and its limited visual semantics. To bridge this gap, the authors propose a multi-domain, multi-order cross-modal knowledge distillation framework that jointly models cross-modal knowledge in both the frequency domain and hypergraph-structured relational space. Specifically, adaptive frequency-domain disentangled feature distillation (AF²D²) enables fine-grained knowledge transfer in the spectral domain, while multi-order relational distillation (MORD) captures higher-order structured semantic relationships. The proposed approach substantially narrows the performance disparity between event-based and frame-based detectors and demonstrates superior robustness and accuracy in complex scenes.

Abstract

Event-based object Detection (EvDet), as a biologically inspired visual perception paradigm, demonstrates superior performance in scenarios demanding high temporal resolution and a wide dynamic range. Nevertheless, the inherent sparse representations and inadequate visual semantics of event data result in a considerable performance disparity between EvDet and frame-based object detection. Previous works attempt to alleviate this cross-modal discrepancy through knowledge distillation, yet they only focus on spatial visual semantics or pair-wise relational information, thus limiting performance in more complex scenarios. To address this challenge, this paper proposes M^2C-EvDet, a Multi-domain and Multi-order Cross-modal knowledge distillation framework for EvDet. Built upon frequency learning and hypergraph computation, M^2C-EvDet integrates two specialized modules: Adaptive Frequency-Decoupled Feature Distillation (AF^2D^2) and Multi-Order Relational Distillation (MORD).

Problem

• Event-based Object Detection
• Cross-modal Knowledge Distillation
• Sparse Representation
• Visual Semantics
• Performance Disparity

Innovation

• Multi-domain Knowledge Distillation
• Multi-order Relational Distillation
• Event-based Object Detection
• Frequency-Decoupled Feature Learning
• Hypergraph Computation