📡 SignalSight – Signal Classification System

SignalSight is an end-to-end intelligent signal classification system built to identify and interpret digital modulation types using deep learning and real-time visualization. It integrates a trained hierarchical model with a user-friendly Flask web interface that allows anyone to upload wireless signal data and view classification results with interpretability.

It automates complex modulation recognition by converting 1D I/Q signals into 2D spectrograms and passing them through a multi-stage classifier pipeline. The system distinguishes between pure, mixed, and jammed/spoofed signals while visualizing critical spectrogram regions that influenced each decision.

---

📌 Table of Contents

• ✨ Key Features
• 🧩 Tech Stack
• 🧠 Trained Models
• 🚀 How to Run the Project
• 📈 Project Architecture
• 🏭 Industrial Applications
• 🌐 User Interface Highlights

---

✨ Key Features

• 🎯 Converts 1D signal arrays into 2D spectrograms for model inference
• 🔁 Hierarchical classification pipeline (pure, mixed, and spoofed signals)
• 🧠 Specialized CNN trained for various SNR levels
• 🖥️ Flask-based web UI for easy real-time testing and visualization
• 🔍 Interpretable AI via Grad-CAM — highlights decisive spectrogram regions
• ⚙️ Supports classification across 24 digital modulation types (QAM, PSK, AM, FM, etc.)

---

🧩 Tech Stack

Component	Technology
Programming	Python
Backend Framework	Flask
Deep Learning	PyTorch
Signal Processing	Matplotlib
Visualization	Grad-CAM
Dataset	RadioML 2018.01A
UI Frontend	HTML/CSS

---

Model Workflow

---

🧠 Trained Models

Pre-trained .pth model files located in checkpoints/:

• best_pure_cnn_model.pth – Pure signal classifier
• best_pure_mixed_model.pth – Mixed signal classifier
• best_qam_specialist_model.pth – QAM modulation specialist
• best_phase_specialist_model.pth – Phase modulation specialist
• best_analog_specialist_model.pth – Analog modulation specialist
• best_jamming_classifier_model.pth – Spoofing and jamming detector
• best_router_model.pth – SNR routing model

---

🚀 How to Run the Project

1️⃣ Clone, Install & Start (single code block)

# Clone the repository
git clone https://github.com/your-username/Signal-Sight.git
cd Signal-Sight

# Install dependencies
pip install -r requirements.txt

# Start the Flask server
python app.py

2️⃣ Verify Model Files

Ensure all .pth model files are present inside the checkpoints/ directory:

• best_pure_cnn_model.pth
• best_pure_mixed_model.pth
• best_qam_specialist_model.pth
• best_phase_specialist_model.pth
• best_analog_specialist_model.pth
• best_jamming_classifier_model.pth
• best_router_model.pth

3️⃣ Access the Web Interface

Open your browser and go to:
http://127.0.0.1:5500/Signal-Sight/templates/index.html

4️⃣ Classify a Signal

• Upload your I/Q signal data in JSON format.
• Click Classify Signal.
• View the predicted modulation type plus the Grad-CAM heatmap for interpretability.