GeminiFusion: Efficient Pixel-wise Multimodal Fusion for Vision Transformer

Benchmark Model Rank Results
semantic-segmentation-on-nyu-depth-v2GeminiFusion (Swin-Large)#1Mean IoU: 60.9
semantic-segmentation-on-nyu-depth-v2GeminiFusion (Swin-Large)#3Mean IoU: 60.2
semantic-segmentation-on-nyu-depth-v2GeminiFusion (MiT-B5)#11Mean IoU: 57.7
semantic-segmentation-on-nyu-depth-v2GeminiFusion (MiT-B3)#17Mean IoU: 56.8
semantic-segmentation-on-deliver-1GeminiFusion#2mIoU: 66.9
semantic-segmentation-on-sun-rgbdGeminiFusion (Swin-Large)#2Mean IoU: 54.6
semantic-segmentation-on-sun-rgbdGeminiFusion (MiT-B5)#7Mean IoU: 53.3
semantic-segmentation-on-sun-rgbdGeminiFusion (MiT-B3)#14Mean IoU: 52.7
semantic-segmentation-on-deliverGeminiFusion#3mIoU: 66.9