MobileCLIP: Fast Image-Text Models through Multi-Modal Reinforced Training
Paper
•
2311.17049
•
Published
MobileCLIP Models + DataCompDR Data
MobileCLIP: Mobile-friendly image-text models with SOTA zero-shot capabilities. DataCompDR: Improved datasets for training image-text SOTA models.
-
-
apple/mobileclip_s0_timm
Image Classification • Updated • 153 • 3 -
apple/mobileclip_s1_timm
Image Classification • Updated • 106 • 1 -
apple/mobileclip_s2_timm
Image Classification • Updated • 160 -
apple/mobileclip_b_timm
Image Classification • Updated • 40 • 1
apple/mobileclip_b_lt_timm
Image Classification • Updated • 66 • 1Note ^ MobileCLIP checkpoints for the timm library (image-tower only)
-
apple/MobileCLIP-S1-OpenCLIP
Zero-Shot Image Classification • Updated • 123 • 4 -
apple/MobileCLIP-S2-OpenCLIP
Zero-Shot Image Classification • Updated • 214 -
apple/MobileCLIP-B-OpenCLIP
Zero-Shot Image Classification • Updated • 30
apple/MobileCLIP-B-LT-OpenCLIP
Zero-Shot Image Classification • Updated • 65 • 2Note ^ MobileCLIP checkpoints for the OpenCLIP library
-
apple/MobileCLIP-S0
Updated • 3 • 2 -
apple/MobileCLIP-S1
Updated • 1 -
apple/MobileCLIP-S2
Updated -
apple/MobileCLIP-B
Updated
apple/MobileCLIP-B-LT
Updated • 2 • 1Note ^ MobileCLIP checkpoints, original format
-
apple/DataComp-12M
Viewer • Updated • 12.8M • 2 • 1 -
apple/DataCompDR-12M
Viewer • Updated • 12.8M • 8 • 11 -
apple/DataCompDR-12M-bf16
Preview • Updated • 22
apple/DataCompDR-1B
Viewer • Updated • 50.3k • 12 • 9Note ^ DataCompDR datasets