MobileCLIP: Fast Image-Text Models through Multi-Modal Reinforced Training

Benchmark Model Rank Results
zero-shot-image-classification-on-imagenet-oMobileCLIP-S0#1Accuracy: 53.8
zero-shot-cross-modal-retrieval-on-flickr30kMobileCLIP-B (LT)#7Image-to-text R@1: 92.3
zero-shot-cross-modal-retrieval-on-flickr30kMobileCLIP-B#26Text-to-image R@1: 77.3
zero-shot-transfer-image-classification-on-imagenet-rMobileCLIP-B#11Accuracy: 89.6
zero-shot-cross-modal-retrieval-on-coco-2014MobileCLIP-B#13Image-to-text R@1: 68.8
zero-shot-transfer-image-classification-on-imagenet-aMobileCLIP-B#13Accuracy (Private): 58.7
image-classification-on-objectnetMobileCLIP-B#20Top-1 Accuracy: 69.4
image-classification-on-imagenet-v2MobileCLIP-B#41Top 1 Accuracy: 69.8