UniRGB-IR: A Unified Framework for RGB-Infrared Semantic Tasks via Adapter Tuning

Benchmark Model Rank Results
multispectral-object-detection-on-kaistUniRGB-IR#4All Miss Rate: 25.21
multispectral-object-detection-on-flirUniRGB-IR#5mAP50: 81.4%mAP: 44.1%
pedestrian-detection-on-llvipUniRGB-IR#9AP: 0.632
thermal-image-segmentation-on-mfn-datasetUniRGB-IR#9mIOU: 59.3
thermal-image-segmentation-on-pst900UniRGB-IR#10mIoU: 82.8