Multi-label Cluster Discrimination for Visual Representation Learning

Open paper
Benchmark Model Rank Results
linear-probe-image-classification-on-dtdMLCD ViT-L/14 (linear probe)#1Accuracy: 84.6
referring-expression-segmentation-on-refcoco-plus-test-bMLCD-Seg-7B#1Overall IoU: 75.6
referring-expression-segmentation-on-refcoco-plus-valMLCD-Seg-7B#1Overall IoU: 79.4
linear-probe-image-classification-on-caltech-101MLCD ViT-L/14 (linear probe)#2Accuracy: 97.2
linear-probe-image-classification-on-fgvc-aircraftMLCD ViT-L/14 (linear probe)#2Accuracy: 79.6
linear-probe-image-classification-on-sun397MLCD ViT-L/14 (linear probe)#2Accuracy: 82.1
referring-expression-segmentation-on-refcoco-plus-testaMLCD-Seg-7B#2Overall IoU: 82.9
referring-expression-segmentation-on-refcocog-testMLCD-Seg-7B#2Overall IoU: 80.5
referring-expression-segmentation-on-refcocog-valMLCD-Seg-7B#2Overall IoU: 79.9
zero-shot-transfer-image-classification-on-fgvc-aircraftMLCD ViT-L/14#2Accuracy: 39.4
linear-probe-image-classification-on-food-101MLCD ViT-L/14 (linear probe)#3Accuracy: 94.3
referring-expression-segmentation-on-refcoco-testaMLCD-Seg-7B#3Overall IoU: 85.3
referring-expression-segmentation-on-refcoco-testbMLCD-Seg-7B#3Overall IoU: 81.5
zero-shot-transfer-image-classification-on-caltech-101MLCD ViT-L/14#3Accuracy: 93.9
zero-shot-transfer-image-classification-on-sunMLCD ViT-L/14#4Accuracy: 72.1
referring-expression-segmentation-on-refcoco-valMLCD-Seg-7B#6Overall IoU: 83.6
zero-shot-transfer-image-classification-on-imagenet-sketchMLCD ViT-L/14#7Accuracy (Private): 60.4
zero-shot-transfer-image-classification-on-food-101MLCD ViT-L/14#10Top 1 Accuracy: 90.3
zero-shot-transfer-image-classification-on-imagenet-rMLCD ViT-L/14#14Accuracy: 85.1
image-classification-on-flowers-102MLCD ViT-L/14 (linear probe)#17Accuracy: 99.3
zero-shot-transfer-image-classification-on-imagenet-aMLCD ViT-L/14#17Accuracy (Private): 56.4
image-classification-on-cifar-100MLCD ViT-L/14 (linear probe)#21Percentage correct: 92.0
zero-shot-cross-modal-retrieval-on-coco-2014MLCD ViT-L/14#21Image-to-text R@1: 60.8
image-classification-on-objectnetMLCD ViT-L/14#24Top-1 Accuracy: 62.7
visual-question-answering-on-docvqa-testMLCD-Embodied-7B#34ANLS: 0.916
image-classification-on-cifar-10MLCD ViT-L/14 (linear probe)#36Percentage correct: 98.9
image-classification-on-imagenet-v2MLCD ViT-L/14#43Top 1 Accuracy: 68.9
image-classification-on-imagenetMLCD ViT-L/14 (finetune)#116Top 1 Accuracy: 87.1%
image-classification-on-imagenetMLCD ViT-L/14 (linear probe)#328Top 1 Accuracy: 84.6%

Rank counts only results with a code link.