AIbase
EN
Home
Vit Base Patch32 Clip 224.laion400m E31
V

Vit Base Patch32 Clip 224.laion400m E31

Developed by timm
Vision Transformer model trained on the LAION-400M dataset, compatible with both OpenCLIP and timm frameworks
Image Classification
Safetensors
Open Source License:MIT#Zero-shot Image Classification#Multimodal CLIP Architecture#LAION-400M Pretraining
Downloads 10.90k
Release Time : 10/23/2024

Model Overview

This is a vision-language model based on the Vision Transformer architecture, primarily used for zero-shot image classification tasks. The model employs 32x32 patch size and 224x224 input resolution, optimized with quickgelu activation function during training.

Model Features

Dual Framework Compatibility
Supports both OpenCLIP and timm frameworks, offering flexible usage options
Quick Activation Function
Utilizes quickgelu activation function to optimize the training process
Large-scale Training Data
Trained on the extensive LAION-400M dataset

Model Capabilities

Zero-shot Image Classification
Image Feature Extraction
Cross-modal Representation Learning

Use Cases

Computer Vision
Image Classification
Classify images of new categories without specific training
Image Retrieval
Retrieve relevant images based on text descriptions
Multimodal Applications
Image-Text Matching
Evaluate the alignment between images and text descriptions
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
EnglishįŽ€äŊ“中文įšéĢ”中文ãĢãģんご
Š 2025AIbase