vit-base-patch16-224-in21k Open-source Handwriting Recognition Model - Free Deployment for Accurate Handwriting Recognition

Home

Vit Base Patch16 224 In21k Writer Identification

Developed by Imene

Fine-tuned based on Google's ViT model for handwriting recognition tasks

Image Classification

Transformers

Open Source License:Apache-2.0 #Handwriting Recognition #ViT Fine-tuning #Multi-classification Model

Downloads 21

Release Time : 9/7/2022

Model Overview

This model is a fine-tuned version of Google's ViT (Vision Transformer) base model for handwriting recognition tasks, capable of identifying different authors' handwriting styles.

Model Features

Transformer-based Vision Model

Uses Vision Transformer architecture to process image data with powerful feature extraction capabilities

Handwriting Recognition Capability

Fine-tuned specifically for handwriting recognition tasks, able to distinguish writing styles of different authors

Transfer Learning

Fine-tuned based on a pre-trained ViT model, leveraging the advantages of large-scale image pre-training

Model Capabilities

Handwriting feature extraction

Author identification

Image classification

Use Cases

Document Analysis

Historical Document Author Identification

Identify potential authors of historical documents or manuscripts

Validation set accuracy 42.55%, Top-3 accuracy 68.84%

Handwriting Verification

Verify whether two documents were written by the same author

Security Verification

Signature Verification

Verify the authenticity of signatures

🚀 Imene/vit-base-patch16-224-in21k-Writer-Identification

This model is a fine - tuned version of a pre - trained model, used for writer identification.

🚀 Quick Start

This model is a fine-tuned version of google/vit-base-patch16-224-in21k on an unknown dataset. It achieves the following results on the evaluation set:

Train Loss: 2.9065
Train Accuracy: 0.6173
Train Top-3-accuracy: 0.8384
Validation Loss: 3.0602
Validation Accuracy: 0.4255
Validation Top-3-accuracy: 0.6884
Epoch: 10

📚 Documentation

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

🔧 Technical Details

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

optimizer: {'inner_optimizer': {'class_name': 'AdamWeightDecay', 'config': {'name': 'AdamWeightDecay', 'learning_rate': {'class_name': 'PolynomialDecay', 'config': {'initial_learning_rate': 1e-05, 'decay_steps': 1500, 'end_learning_rate': 0.0, 'power': 1.0, 'cycle': False, 'name': None}}, 'decay': 0.0, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-08, 'amsgrad': False, 'weight_decay_rate': 0.01}}, 'dynamic': True, 'initial_scale': 32768.0, 'dynamic_growth_steps': 2000}
training_precision: mixed_float16

Training results

Train Loss	Train Accuracy	Train Top-3-accuracy	Validation Loss	Validation Accuracy	Validation Top-3-accuracy	Epoch
3.8889	0.0401	0.0971	3.8524	0.0501	0.1577	0
3.7785	0.1516	0.3210	3.7267	0.1464	0.3392	1
3.6188	0.2656	0.4854	3.5946	0.2115	0.4143	2
3.4631	0.3276	0.5860	3.4570	0.2728	0.5232	3
3.3386	0.3959	0.6599	3.3664	0.3091	0.5757	4
3.2366	0.4500	0.7094	3.2844	0.3342	0.5895	5
3.1510	0.4879	0.7526	3.2364	0.3442	0.6170	6
3.0757	0.5246	0.7823	3.1752	0.3830	0.6508	7
3.0107	0.5612	0.8058	3.1281	0.4043	0.6596	8
2.9549	0.5894	0.8212	3.0903	0.4255	0.6834	9
2.9065	0.6173	0.8384	3.0602	0.4255	0.6884	10

Framework versions

Transformers 4.21.3
TensorFlow 2.8.2
Datasets 2.4.0
Tokenizers 0.12.1

📄 License

This model is licensed under Apache-2.0.

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