GaMi: Multimodal Material ID via mmWave and Acoustic Sensing

Researchers have developed GaMi, a multimodal material identification system that combines mmWave and acoustic sensing to overcome geometry-induced variations and single-modality ambiguities. The system uses an intra-sample cross-modal subtractive disentanglement framework, leveraging shared geometric consistency between co-located bimodal sensors to isolate intrinsic material features. Inter-sample contrastive learning corrects residual interference from cross-modal misalignment, while a pairing-based adaptation strategy enables few-shot generalization across devices. The work, published on arXiv (2605.30818), targets non-contact material identification for embodied intelligence, operating robustly under unconstrained geometric conditions such as varying orientation, shape, and distance.