SmartCLIP: Modular Vision-language Alignment with Identification Guarantees

Benchmark Model Rank Results
zero-shot-transfer-image-classification-on-fgvc-aircraftSmartCLIP (ViT-L/14)#3Accuracy: 30.4
zero-shot-transfer-image-classification-on-sunSmartCLIP (ViT-L/14)#5Accuracy: 72.1
image-classification-on-imagenet-v2SmartCLIP (ViT-L/14)#50Top 1 Accuracy: 66.6