Anti Spoofing System

This is project for exploration about Machine Learning Lifecycle in Classification Image Task

Folder Structure

│
├── Figures                   <- Example Image for input and output
│
│
├── src                       <- Source code
│   ├── data                      <- Data scripts
│   ├── models                    <- Model scripts
│   ├── metrics                   <- Calculation metrics scripts
│   ├── pipelines                 <- Machine learning pipeline for training and evaluation scripts
│   ├── api                       <- Endpoint script
│   ├── services                  <- Service script
│   ├── utils.py                  <- Utility scripts
|
|
│
├── Dockerfile                <- Containerization
├── docker-compose.yml        <- Container Orchestration
├── app.py                    <- Run API Endpoint
├── run_training.py           <- Run Training Pipeline
├── run_evaluation.py         <- Run Evaluation Pipeline
├── predict_sample.py         <- Example code how to predict image
├── .gitignore                <- List of files ignored by git
├── requirements.txt          <- List python dependencies
├── setup.sh                  <- File for set up python dependencies
└── README.md

Getting Started

Prerequisites

• Docker
• Docker Compose

Installation

1. Clone the repository:

   git clone https://github.com/MuhFaridanSutariya/anti-spoofing.git
   cd anti-spoofing

2. Build and run the Docker container:

   docker-compose up --build
   

Usage

The API can be accessed at http://localhost:5000. You can use tools like Postman to interact with the API.

Example Request To make a prediction, you can send a POST request to the /predict endpoint with an image file.

References

• Mobilenet v2 Architecture
• Senet Architecture
• Framework PyTorch

Todo:

• CI/CD using Github Action