vit-base-mnist Open-source Image Classification Model - Fine-tuned on MNIST with an Accuracy of up to 99.49%

Vit Base Mnist

Developed by farleyknight-org-username

An image classification model fine-tuned on the MNIST dataset based on the ViT architecture, achieving an accuracy of 99.49%

Image Classification

Transformers

Open Source License:Apache-2.0 #High-precision image classification #MNIST dataset #ViT architecture

Downloads 1,770

Release Time : 8/21/2022

Model Overview

This model is an image classification model fine-tuned on the MNIST handwritten digit dataset based on Google's ViT-base architecture, specifically designed to recognize handwritten digits from 0 to 9.

Model Features

High accuracy

Achieves 99.49% accuracy on the MNIST test set

Based on ViT architecture

Uses the Vision Transformer (ViT) architecture, which has advantages over traditional CNNs in image classification tasks

Lightweight fine-tuning

Fine-tuned based on a pre-trained model, resulting in high training efficiency

Model Capabilities

Handwritten digit recognition

Image classification

Digit recognition

Use Cases

Education

Handwritten digit recognition system

Used for automatically grading handwritten digit assignments or exams

Recognition accuracy of 99.49%

Finance

Check digit recognition

Automatically recognizes handwritten amount digits on checks

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.3717	1.0	6375	0.0522	0.9893
0.3453	2.0	12750	0.0370	0.9906
0.3736	3.0	19125	0.0308	0.9916
0.3224	4.0	25500	0.0269	0.9939
0.2846	5.0	31875	0.0236	0.9949

Property	Details
Model Type	Fine - tuned Vision Transformer (ViT)
Training Data	mnist
Metrics	Accuracy
Results	Accuracy on mnist train split: 0.9948888888888889

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご

Vit Base Mnist

Model Overview

Model Features

Model Capabilities

Use Cases

🚀 vit-base-mnist

🚀 Quick Start

📚 Documentation

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

📄 License

📦 Model Information