🤖 SETI Signal Classifier

Image classifier built with Convolutional Neural Networks (CNN) and Class Activation Map (CAM) support — a project for classifying radio-signal-style images and visualizing model interpretation during training.

Live demo: https://huggingface.co/spaces/Esabelle/seti_classifier_demo

training notebook: https://www.kaggle.com/code/esabellechen/seti-signal-classifier-vgg16-pytorch https://www.kaggle.com/code/esabellechen/seti-signal-classifier-googlene-pytorch

---

🔍 Overview

This project implements a deep learning-based image classification pipeline using CNNs to classify images and interpret model decisions through Class Activation Maps (CAMs).

It is especially suited for visualizing how the network focuses on areas of the image during training and inference. The notebook and code include:

✔️ CNN model training
✔️ Image preprocessing
✔️ CAM visualization
✔️ Evaluation and plotting

Demo Video

https://youtu.be/yxtqx7oFAUk

---

📂 Project Contents

Folder / File	Description
app/	Source for model training and classification logic
images/test/	Example test images
model/	Saved model checkpoints
seti-image-classification-notebook.ipynb	Notebook for exploration & training
requirements.txt	Python dependencies
Dockerfile	For containerized setup
docker-compose.yml	For running with Docker services

---

🛠️ Features

• 📷 Image Classification using Convolutional Neural Networks
• 🧠 Class Activation Maps (CAM) for visual interpretability
• 📊 Training + evaluation workflow
• 🧪 Test images included
• 🐳 Docker support for repeatable environment

---

🚀 Get Started

Prerequisites

• Python 3.11
• Docker & Docker Compose (optional)
• 4GB+ RAM (for transformer models)

Installation

Local Setup

1. Clone the repository

    git clone https://github.com/Esabelle11/seti_classifier.git
    cd seti_classifier

2. Create and activate virtual environment

   python -m venv venv
   
   # On Windows
   venv\Scripts\activate
   
   # On Linux/Mac
   source venv/bin/activate

3. Install dependencies

   pip install -r requirements.txt

4. Run the application

   uvicorn app.main:app --reload

   The API will be available at http://localhost:8000

Docker Setup

Using Docker Compose (Recommended)

1. Build and run

   docker-compose up -d

2. Access the application

   • Frontend: http://localhost:8000
   • API Docs: http://localhost:8000/docs

3. Stop the application

   docker-compose down

Using Docker Only

1. Build the image

   docker build -t seti-classifier .

2. Run the container

   docker run -p 8888:8888 seti-classifier