Classify Radio Signals from Outer Space with Keras

by brewbooks is licensed under CC BY 2.0

This project demonstrates the process of using a Convolutional Neural Network (CNN) to classify space signals into four categories: "squiggle", "narrowband", "noise", and "narrowbanddrd".

Setup Instructions

• Create a virtual environment.
• Unzip dataset.zip.
• Activate the virtual environment and install dependencies:

.venv\Scripts\Activate.ps1
pip install --upgrade pip ipykernel scikit-learn tensorflow pandas numpy matplotlib seaborn livelossplot
pip freeze > requirements.txt

Training Process

The training involves the following steps:

Step 1: Import Libraries

• Essential libraries and modules are imported.
• TensorFlow's version is printed to ensure compatibility.

Step 2: Load and Preprocess SETI Data

• The SETI dataset, transformed into images, is loaded.
• Data is reshaped to fit the model requirements.

Step 3: Plot 2D Spectrograms

• Spectrograms of the signals are plotted to visualize the data.

Step 4: Create Training and Validation Data Generators

• ImageDataGenerators are used for data augmentation to improve model robustness.

Step 5: CNN Architecture

• The CNN model is defined with layers designed to capture spatial hierarchies in the data.

Step 6: Schedule Learning Rate and Compile Model

• An exponential decay learning rate schedule is implemented.
• The model is compiled with the Adam optimizer.

Step 7: Callbacks

• Callbacks including model checkpointing and plot losses are set up to monitor training.

Step 8: Model Training

• The model is trained using the defined data generators and callbacks.

Step 9: Model Evaluation, Prediction, Classification Report

• The model's performance is evaluated on a validation set.
• A classification report and confusion matrix provide detailed performance metrics.

Conclusion

This project outlines a comprehensive approach to classifying radio signals from space using a CNN with Keras. It covers the end-to-end process, from data preprocessing and model definition to training, evaluation, and interpretation of results.