Universal Instance Perception as Object Discovery and Retrieval

Benchmark Model Rank Results
multi-object-tracking-and-segmentation-on-bdd100k-valUNINEXT-H#1mMOTSA: 35.7
referring-expression-segmentation-on-davis-2017-valUNINEXT-H#1J&F 1st frame: 72.5
visual-tracking-on-tnl2kUNINEXT-H#2AUC: 59.3precision: 62.8
generalized-referring-expression-comprehension-on-grefcocoUNINEXT#3Precision@(F1=1, IoU≥0.5): 58.2N-acc.: 50.6
multiple-object-tracking-on-bdd100k-valUNINEXT-H#3mMOTA: 44.2mIDF1: 56.7
zero-shot-segmentation-on-segmentation-in-the-wildUNINEXT#3Mean AP: 42.1
described-object-detection-on-description-detectionUNINEXT-large#5Intra-scenario FULL mAP: 17.9Intra-scenario PRES mAP: 18.6…
visual-object-tracking-on-lasot-extUNINEXT-H#6AUC: 56.2Normalized Precision: 63.8Precision: 63.8
referring-expression-segmentation-on-refcoco-plus-test-bUNINEXT-H#9Overall IoU: 66.22
referring-expression-segmentation-on-refcoco-valUNINEXT-H#11Overall IoU: 82.19
video-instance-segmentation-on-ovis-validationUNINEXT (ViT-H, Online)#11mask AP: 49.0AP50: 72.5AP75: 52.2
video-instance-segmentation-on-ovis-validationUNINEXT (ResNet-50, Online)#29mask AP: 34.0AP50: 55.5AP75: 35.6
referring-expression-segmentation-on-refcoco-plus-testaUNINEXT-H#12Overall IoU: 76.42
referring-expression-segmentation-on-refcoco-plus-valUNINEXT-H#13Overall IoU: 72.47
referring-expression-segmentation-on-refer-youtube-vos-2021-public-validationUNINEXT-H#13J&F: 70.1J: 67.6F: 72.7
instance-segmentation-on-coco-test-devUNINEXT-H#16mask AP: 51.8AP50: 76.2AP75: 56.7APS: 33.3APM: 55.9APL: 67.5
visual-object-tracking-on-trackingnetUNINEXT-H#16Accuracy: 85.4Normalized Precision: 89.0Precision: 86.4
object-detection-on-coco-val2017UNINEXT-H#23box AP: 60.6AP50: 77.5AP75: 66.7APS: 45.1APM: 64.8APL: 75.3
visual-object-tracking-on-lasotUNINEXT-L#23AUC: 72.4Normalized Precision: 80.7Precision: 78.9
visual-object-tracking-on-lasotUNINEXT-H#26AUC: 72.2Normalized Precision: 80.8Precision: 79.4