The paper "EventDance: Unsupervised Source-free Cross-modal Adaptation for Event-based Object Recognition" addresses the challenging task of cross-modal adaptation from image to event modalities without access to labeled source image data. The authors propose a novel framework called EventDance, which includes two main modules: Reconstruction-based Modality Bridging (RMB) and Multi-representation Knowledge Adaptation (MKA). RMB constructs surrogate images from events to bridge the modality gap, while MKA transfers knowledge from the source model to target models using multiple event representations. The framework is evaluated on three event-based recognition benchmarks, demonstrating superior performance compared to prior methods. The key contributions include addressing a novel and challenging problem, proposing EventDance, and showing its effectiveness through extensive experiments.The paper "EventDance: Unsupervised Source-free Cross-modal Adaptation for Event-based Object Recognition" addresses the challenging task of cross-modal adaptation from image to event modalities without access to labeled source image data. The authors propose a novel framework called EventDance, which includes two main modules: Reconstruction-based Modality Bridging (RMB) and Multi-representation Knowledge Adaptation (MKA). RMB constructs surrogate images from events to bridge the modality gap, while MKA transfers knowledge from the source model to target models using multiple event representations. The framework is evaluated on three event-based recognition benchmarks, demonstrating superior performance compared to prior methods. The key contributions include addressing a novel and challenging problem, proposing EventDance, and showing its effectiveness through extensive experiments.