The proposed Fusion-Mamba method effectively fuses features from different modalities (RGB and infrared) by associating them in a hidden state space, reducing disparities between cross-modal features and enhancing the representation consistency of fused features.
서로 다른 모달리티의 특징을 효과적으로 융합하여 객체 탐지 성능을 향상시키는 Fusion-Mamba 방법을 제안한다.