Universal Instance Perception as Object Discovery and Retrieval

Benchmark Model Rank Results
described-object-detection-on-descriptionUNINEXT-large#5Intra-scenario FULL mAP: 17.9Intra-scenario PRES mAP: 18.6
generalized-referring-expressionUNINEXT#2Precision@(F1=1, IoU≥0.5): 58.2N-acc.: 50.6
instance-segmentation-on-cocoUNINEXT-H#16mask AP: 51.8AP50: 76.2AP75: 56.7APS: 33.3APM: 55.9APL: 67.5
multi-object-tracking-and-segmentation-on-3UNINEXT-H#1mMOTSA: 35.7
multiple-object-tracking-on-bdd100k-valUNINEXT-H#2mMOTA: 44.2mIDF1: 56.7
object-detection-on-coco-minivalUNINEXT-H#22box AP: 60.6AP50: 77.5AP75: 66.7APS: 45.1APM: 64.8APL: 75.3
referring-expression-segmentation-on-davisUNINEXT-H#1J&F 1st frame: 72.5
referring-expression-segmentation-on-refcocoUNINEXT-H#7Overall IoU: 82.19
referring-expression-segmentation-on-refcoco-3UNINEXT-H#11Overall IoU: 72.47
referring-expression-segmentation-on-refcoco-4UNINEXT-H#9Overall IoU: 76.42
referring-expression-segmentation-on-refcoco-5UNINEXT-H#8Overall IoU: 66.22
referring-expression-segmentation-on-refer-1UNINEXT-H#4J&F: 70.1J: 67.6F: 72.7
video-instance-segmentation-on-ovis-1UNINEXT (ViT-H, Online)#7mask AP: 49.0AP50: 72.5AP75: 52.2
video-instance-segmentation-on-ovis-1UNINEXT (ResNet-50, Online)#25mask AP: 34.0AP50: 55.5AP75: 35.6
visual-object-tracking-on-lasotUNINEXT-L#19AUC: 72.4Normalized Precision: 80.7Precision: 78.9
visual-object-tracking-on-lasotUNINEXT-H#21AUC: 72.2Normalized Precision: 80.8Precision: 79.4
visual-object-tracking-on-lasot-extUNINEXT-H#5AUC: 56.2Normalized Precision: 63.8Precision: 63.8
visual-object-tracking-on-trackingnetUNINEXT-H#13Accuracy: 85.4Normalized Precision: 89.0Precision: 86.4
visual-tracking-on-tnl2kUNINEXT-H#2AUC: 59.3precision: 62.8
zero-shot-segmentation-on-segmentation-in-theUNINEXT#3Mean AP: 42.1