Cross-modality attentive feature fusion for object detection in multispectral remote sensing imagery

Cross-modality fusing complementary information of multispectral remote sensing image pairs can improve the perception ability detection algorithms, making them more robust and reliable for a wider range applications, such as nighttime detection. Compared with prior methods, we think different features should be processed specifically, modality-specific retained enhanced, while modality-shared cherry-picked from RGB thermal IR modalities. Following this idea, novel lightweight feature fusion approach joint common-modality differential-modality attentions are proposed, named Cross-Modality Attentive Feature Fusion (CMAFF). Given intermediate maps images, our module parallel infers attention two separate modalities, common- differential-modality, then multiplied to input map respectively adaptive enhancement or selection. Extensive experiments demonstrate that proposed achieve state-of-the-art performance at low computation cost.