iBOT: Image BERT Pre-Training with Online Tokenizer

Open paper
Benchmark Model Rank Results
instance-segmentation-on-cocoiBOT (ViT-B/16)#43mask AP: 44.2
instance-segmentation-on-cocoiBOT (ViT-S/16)#49mask AP: 42.6
object-detection-on-cocoiBOT (ViT-B/16)#83box mAP: 51.2
object-detection-on-cocoiBOT (ViT-S/16)#95box mAP: 49.4
self-supervised-image-classification-on-1iBOT(ViT-L/16, 512)#8Top 1 Accuracy: 87.8%Number of Params: 307M
self-supervised-image-classification-on-1iBOT(ViT-L/16)#12Top 1 Accuracy: 86.6%Number of Params: 307M
self-supervised-image-classification-on-1iBOT (ViT-L/16)#26Top 1 Accuracy: 84.8%Number of Params: 307M
self-supervised-image-classification-on-1iBOT (ViT-B/16)#31Top 1 Accuracy: 84.4%Number of Params: 85M
self-supervised-image-classification-on-1iBOT (ViT-B/16)#38Top 1 Accuracy: 84.0%Number of Params: 85M
self-supervised-image-classification-on-imagenetiBOT (ViT-L/16) (IN22k)#11Top 1 Accuracy: 82.3%Number of Params: 307M
self-supervised-image-classification-on-imagenetiBOT (ViT-L/16)#16Top 1 Accuracy: 81.3%Number of Params: 307M
semantic-segmentation-on-ade20kiBOT (ViT-B/16)#119Validation mIoU: 50.0
semantic-segmentation-on-ade20kiBOT (ViT-S/16)#184Validation mIoU: 45.4
semantic-segmentation-on-ade20kiBOT (ViT-B/16) (linear head)#213Validation mIoU: 38.3
semi-supervised-image-classification-on-1iBOT (ViT-S/16)#32Top 1 Accuracy: 61.9%
unsupervised-image-classification-on-imagenetiBOT (ViT-S/16)#3Accuracy (%): 43.4ARI: 32.8

Rank counts only results with a code link.