---

PMDL Assignment №1: MLOps Solution for X-Ray Chest Kaggle Competition

This repository contains the MLOps pipeline developed by Ivan Golov, a student of Innopolis University (AI-01 Group). The goal of this project is to build and deploy a machine learning model to classify chest X-ray images as part of a Kaggle competition.

Overview

In this repository, you'll find a basic machine learning model that processes X-ray chest images and performs classification. The model is built and trained using Jupyter notebooks, which are available in the notebooks/ directory.

After model training and validation, the solution is deployed using Gradio as the front-end interface and Docker for containerization, making it easy to deploy in any environment.

Key Components

1. Model Development:

   • The basic machine learning model is implemented using standard libraries and frameworks. Please refer to the notebooks in the notebooks/ folder for details on model architecture, training procedures, and evaluation metrics.

2. Gradio Interface:

   • A user-friendly interface is built using Gradio to allow users to upload X-ray images and receive predictions directly from the model.

3. Docker Deployment:

   • The project is containerized using Docker for seamless deployment. Instructions for building and running the Docker container can be found in the Dockerfile and below.

Setup Instructions

Local Setup

1. Clone the repository:

   git clone https://github.com/IVproger/PMDL_MLops.git
   cd PMDL_MLops

2. Install Dependencies: Make sure you have Python and the required libraries installed. If you want run notebooks and modify the project, you can install dependencies using the provided requirements.txt:

   pip install -Ur requirements.txt

Docker Setup

1. Model installation

   Make sure that before the building of docker images, you installed the model pth file and put it inside the deployment/models folder.

2. Build Docker Image:

   docker-compose build
   

3. Run Docker Container:

   docker-compose up
   

   This will start the Gradio app on http://localhost:7860.

Contact

For any questions or contributions, feel free to reach out:

• Ivan Golov
• Email: i.golov@innopolis.university

---