Open-source model for facial age image detection - Accurately predict age ranges based on facial images

Home

Facial Age Image Detection

Developed by dima806

A model trained using Vision Transformer (ViT) architecture to predict age ranges from facial images

Face-related

Transformers

Open Source License:Apache-2.0 #Facial Age Range Recognition #ViT Vision Transformer #Multi-age Classification

Downloads 768

Release Time : 11/4/2023

Model Overview

This model is fine-tuned from the google/vit-base-patch16-224-in21k pre-trained model and can identify and classify 23 different age ranges from facial images. Suitable for scenarios such as demographic analysis and age verification.

Model Features

Multi-age Range Classification

Capable of recognizing 23 fine-grained age ranges from 0-1 years to 90+

ViT-based Architecture

Utilizes Vision Transformer model, excelling in image classification tasks

High Accuracy

Achieves F1 scores above 90% for certain age ranges (e.g., 0-1 years and 90+)

Model Capabilities

Facial Image Analysis

Age Range Prediction

Population Age Distribution Analysis

Use Cases

Demographic Analysis

Mall Customer Age Analysis

Analyzing customer age distribution via surveillance cameras

Helps businesses understand the main customer age groups

Digital Identity Verification

Age Verification System

Verifying if users meet age requirements for specific services

Prevents minors from accessing inappropriate content

🚀 Facial Age Detection Model

This project provides a model that can return the age bin based on a facial image, offering a solution for age classification tasks in facial recognition scenarios.

✨ Features

Metrics: The model is evaluated using metrics such as accuracy and F1-score.
Base Model: It is built on the google/vit-base-patch16-224-in21k model.

📦 Information

Property	Details
Model Type	Classification model for facial age detection
Base Model	`google/vit-base-patch16-224-in21k`
Metrics	Accuracy, F1
License	Apache-2.0

💻 Usage Examples

For more detailed usage examples and implementation details, please refer to this Kaggle notebook.

📚 Documentation

Classification Report

The following is the classification report of the model:

Classification report:

              precision    recall  f1-score   support

          01     0.7341    0.9056    0.8109       445
          02     0.4494    0.6787    0.5407       445
          03     0.6978    0.2854    0.4051       445
          04     0.8421    0.1438    0.2457       445
          05     0.5707    0.9618    0.7163       445
       06-07     0.7030    0.5798    0.6355       445
       08-09     0.6500    0.8180    0.7244       445
       10-12     0.6993    0.7056    0.7025       445
       13-15     0.8034    0.7438    0.7725       445
       16-20     0.7006    0.7416    0.7205       445
       21-25     0.6796    0.6292    0.6534       445
       26-30     0.4241    0.5843    0.4915       445
       31-35     0.4654    0.2270    0.3051       445
       36-40     0.4606    0.3416    0.3923       445
       41-45     0.5074    0.6944    0.5863       445
       46-50     0.4896    0.5811    0.5314       444
       51-55     0.5158    0.5506    0.5326       445
       56-60     0.5000    0.3491    0.4111       444
       61-65     0.7083    0.1910    0.3009       445
       66-70     0.4778    0.7995    0.5981       444
       71-80     0.7687    0.7169    0.7419       445
       81-90     0.8425    0.9978    0.9136       445
         90+     0.9978    1.0000    0.9989       444

    accuracy                         0.6185     10231
   macro avg     0.6386    0.6185    0.5970     10231
weighted avg     0.6386    0.6185    0.5970     10231

Model Visualization

📄 License

This project is licensed under 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