vit_Liveness_detection_v1.0 Open-source Facial Liveness Detection Model - Excellent performance on the evaluation set, free to use!

Vit Liveness Detection V1.0

Developed by nguyenkhoa

This model is a face liveness detection model based on the Transformer library and has achieved excellent performance on the evaluation set.

Face-related

Transformers

Open Source License:Apache-2.0 #High-precision live detection #ViT fine-tuning model #Financial-grade security verification

Downloads 176

Release Time : 1/8/2025

Model Overview

This model is a face liveness detection model fine-tuned on google/vit-base-patch16-224, mainly used to distinguish real faces from forged faces.

Model Features

High accuracy

Achieved an accuracy, F1 score, recall rate, and precision rate of 0.9988 on the evaluation set

Based on the ViT architecture

Uses the Vision Transformer architecture, which has powerful feature extraction capabilities

Efficient training

Adopts automatic mixed-precision training, with high training efficiency

Model Capabilities

Face liveness detection

Image classification

Forged face recognition

Use Cases

Identity verification

Financial identity verification

Used in identity verification systems for scenarios such as banks and payments

Effectively prevent forgery attacks such as photos and videos

Access control system

Used for face liveness detection in intelligent access control

Improve the security of the access control system

🚀 vit_Liveness_detection_v1.0

This is a fine - tuned model based on the Vision Transformer architecture, designed for liveness detection tasks.

🚀 Quick Start

This model is a fine - tuned version of [google/vit - base - patch16 - 224](https://huggingface.co/google/vit - base - patch16 - 224) on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.0047
Accuracy: 0.9988
F1: 0.9988
Recall: 0.9988
Precision: 0.9988

📚 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:

learning_rate: 5e - 05
train_batch_size: 128
eval_batch_size: 8
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon = 1e - 08
lr_scheduler_type: linear
num_epochs: 5
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Recall	Precision
0.0254	0.2048	128	0.0148	0.9946	0.9946	0.9946	0.9946
0.0256	0.4096	256	0.0180	0.9945	0.9944	0.9945	0.9945
0.0113	0.6144	384	0.0133	0.9955	0.9955	0.9955	0.9955
0.0116	0.8192	512	0.0070	0.9976	0.9976	0.9976	0.9976
0.0084	1.024	640	0.0072	0.9976	0.9976	0.9976	0.9976
0.0048	1.2288	768	0.0084	0.9976	0.9976	0.9976	0.9976
0.0041	1.4336	896	0.0078	0.9975	0.9975	0.9975	0.9975
0.0015	1.6384	1024	0.0049	0.9983	0.9983	0.9983	0.9983
0.0047	1.8432	1152	0.0068	0.9977	0.9977	0.9977	0.9977
0.0012	2.048	1280	0.0075	0.9975	0.9975	0.9975	0.9975
0.0025	2.2528	1408	0.0095	0.9971	0.9971	0.9971	0.9971
0.0013	2.4576	1536	0.0084	0.9976	0.9976	0.9976	0.9976
0.0026	2.6624	1664	0.0056	0.9985	0.9985	0.9985	0.9985
0.0001	2.8672	1792	0.0096	0.9976	0.9976	0.9976	0.9976
0.0001	3.072	1920	0.0049	0.9987	0.9987	0.9987	0.9987
0.0009	3.2768	2048	0.0085	0.9978	0.9978	0.9978	0.9978
0.0003	3.4816	2176	0.0078	0.9980	0.9980	0.9980	0.9980
0.0002	3.6864	2304	0.0057	0.9985	0.9985	0.9985	0.9985
0.0	3.8912	2432	0.0043	0.9988	0.9988	0.9988	0.9988
0.0	4.096	2560	0.0046	0.9987	0.9987	0.9987	0.9987
0.0	4.3008	2688	0.0045	0.9988	0.9988	0.9988	0.9988
0.0	4.5056	2816	0.0046	0.9988	0.9988	0.9988	0.9988
0.0	4.7104	2944	0.0047	0.9988	0.9988	0.9988	0.9988
0.0	4.9152	3072	0.0047	0.9988	0.9988	0.9988	0.9988

Framework versions

Transformers 4.44.2
Pytorch 2.4.1+cu121
Datasets 3.2.0
Tokenizers 0.19.1

📄 License

The project uses the apache - 2.0 license.

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