AI Image classifiers on Django server with REST API

Table of Contents

• Introduction
• Description
• Features
• Architecture
• Usage
• Results
• Issue
• Others
• License

Introduction

A Keras deep learning image classifiers on Django server with REST API. It can help you quickly deploy and apply ML models.

Description

Machine Learning (ML) models are typically trained and test on benchmark data sets. However, such data sets may not represent the data in real applications. These ML models should be updated after tested on real data sets. Deployment of the models after each update may be time consuming. Therefore, automated deployment is needed to save time and effort. In this proposal, we describe our solution to integrate trained ML models into any application of interest. We use Django web-based platform to build REST API for model deployment. Jenkins is used for continuous integration and automated deployment of ML models into Django servers.

Features

• Add Account management funtion
• Add Login & Signup
• Add Django Rest Framework
• Optimize front-end interface, adapt to mobile and PC interface
• Design front-end and back-end interactive interfaces
• Add feature: Face comparison
• Add feature: Bank card identification
• Add feature: Gesture identification
• Add feature: Image classify using following model：
  • ResNet50
  • Xception
  • MobileNet, MobileNetV2
  • InceptionV3, InceptionResNetV2
  • DenseNet121, DenseNet169, DenseNet201
  • VGG16, VGG19
  • NASNetMobile, NASNetLarge
• Run each part of functions seperately in local and cloud server
• Run all the functions in local server and cloud server
• Design and using machine learning CI/CD Tools such as Git, Jenkins, Nginx, uwgsi for deployment.

Architicture

Usage

I assume you already have your own local virtual environment.

git clone https://github.com/nature1995/image-classify-django-server.git

pip install -r requirements.txt
python manage.py makemigrations
python manage.py migrate
python manage.py collectstatic
python manage.py runserver 0.0.0.0:8000

Access the web page though this link: http://127.0.0.1:8000/

Docker

You can use dockerfile to build this project easily.

# Follow any instruction to install the docker-ce, docker for OS X or PC in your device
docker build -t ai_image_classifiers .
docker run -it --rm -p 8000:8000 --name ai_image_classifiers_app ai_image_classifiers:latest

Result

Input

Note: You can not send a very large size image.

Result

Accuracy for individual models

Example

Using Postman to test the API:
POST http://127.0.0.1:8000/predict_VGG19/api/

Result

{
    "success": true,
    "predictions": [
        {
            "label": "red_fox",
            "probability": 0.8969062566757202
        },
        {
            "label": "kit_fox",
            "probability": 0.08841043710708618
        },
        {
            "label": "grey_fox",
            "probability": 0.012036639265716076
        },
        {
            "label": "Arctic_fox",
            "probability": 0.0022438077721744776
        },
        {
            "label": "coyote",
            "probability": 0.0002566342300269753
        },
        {
            "label": "white_wolf",
            "probability": 0.00005685776341124438
        }
    ]
}

Compatibility

Support to Django>=2.1.5 or Django==3.0.0 and Python 3.5, 3.6, 3.7.

Issue

If you have questions or issues, please feel free to tell us.

Contribution

Welcome to make pull request. If you have a related project/component/tool, add it with a pull request to add it!

Others

Admin Account

python manage.py createsuperuser

username: ranxiaolang
email: YOUR EMAIL  
password: ranxiaolang  

Access the web page though this link: http://127.0.0.1:8000/admin

Django Restframework

Access the web page though this link: http://127.0.0.1:8000/iot/

Contributors

@nature1995

@sergulaydore

License

MIT ©