Bi-directional Cross-Modality Feature Propagation with Separation-and-Aggregation Gate for RGB-D Semantic Segmentation

Benchmark Model Rank Results
object-detection-on-dsecSAGate#11mAP: 19.6
object-detection-on-pku-ddd17-carSAGate#6mAP50: 82.0
semantic-segmentation-on-bjroadSA-Gate#3IoU: 62.14
semantic-segmentation-on-event-basedSA-Gate#3mIoU: 84.08
semantic-segmentation-on-eventscapeSA-Gate#5mIoU: 53.94
semantic-segmentation-on-llrgbd-syntheticSA-Gate (ResNet-101)#8mIoU: 61.79
semantic-segmentation-on-nyu-depth-v2SA-Gate#37Mean IoU: 52.4%
semantic-segmentation-on-portoSA-Gate#3IoU: 72.21
semantic-segmentation-on-potsdamSA-Gate#6mIoU: 84.28
semantic-segmentation-on-sun-rgbdTokenFusion (Ti)#18Mean IoU: 49.4%
semantic-segmentation-on-tlcgisSA-Gate#1IoU: 84.20
semantic-segmentation-on-urbanlfSA-Gate#4mIoU (Syn): 79.53mIoU (Real): n.a.
semantic-segmentation-on-us3dSA-Gate#2mIoU: 83.62
semantic-segmentation-on-vaihingenSA-Gate#2mIoU: 81.03
thermal-image-segmentation-on-mfn-datasetSA-Gate#39mIOU: 45.8
thermal-image-segmentation-on-noisy-rs-rgb-tSA-Gate#4mIoU: 54.0