AIbase
EN
Home
Metaclip H14 Fullcc2.5b
M

Metaclip H14 Fullcc2.5b

Developed by facebook
MetaCLIP is a vision-language model based on CommonCrawl data, improving CLIP model performance through enhanced data filtering methods
Text-to-Image
Transformers
#Zero-shot image classification#Multimodal embedding space#Trained on 2.5 billion data points
Downloads 26.29k
Release Time : 10/9/2023

Model Overview

This model constructs an image-text shared embedding space using MetaCLIP technology, supporting zero-shot image classification and cross-modal retrieval tasks, aiming to address the opacity in CLIP training data filtering

Model Features

Enhanced data filtering method
Optimizes CLIP training data selection process through MetaCLIP technology, addressing the opacity in original CLIP data preprocessing
High-resolution processing capability
Supports 14×14 image patch resolution, capturing finer visual features
Large-scale pretraining
Trained on 2.5 billion data points from CommonCrawl, with strong generalization capabilities

Model Capabilities

Zero-shot image classification
Text-based image retrieval
Image-based text retrieval
Cross-modal feature extraction

Use Cases

Content retrieval
Multimodal search engine
Retrieve relevant images using text queries, or find related text content through images
Intelligent annotation
Automatic image tagging
Generate descriptive labels for unannotated images
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
English简体中文繁體中文にほんご
© 2025AIbase