CIFAR-10 Classification with Fast ML

Objective

This notebook demonstrates the end-to-end process of building and deploying a machine learning model for image classification on the CIFAR-10 dataset using the Fast ML library. The notebook covers data ingestion, preprocessing, model training, evaluation, and deployment.

Problem Definition

We are dealing with an image classification problem where the goal is to classify images into one of the ten classes in the CIFAR-10 dataset. This type of problem is common in computer vision applications.

Pipeline Overview

1. Data Ingestion: Load data from a MongoDB database using the MongoDBConnector.
2. Data Preprocessing: Perform image augmentation, normalization, and resizing.
3. Model Training: Train a deep learning model using TensorFlow/Keras.
4. Model Evaluation: Evaluate the model using accuracy and other relevant metrics.
5. Model Deployment: Deploy the trained model using Docker and MLflow.
6. Experiment Tracking: Track experiments and model performance using MLflow.

Tools and Technologies

• Data Ingestion: MongoDBConnector from Fast ML's data ingestion module.
• Data Preprocessing: Utilize Fast ML's preprocessing functions for images.
• Model Training and Evaluation: Use TensorFlow/Keras for modeling and evaluating performance.
• Deployment: Docker for containerization and MLflow for managing the model's lifecycle.
• Experiment Tracking: MLflow for logging experiments, model parameters, and metrics.