NeuralForge

A visual neural network training platform built with React + PyTorch. Design, train, and test CNN models through an intuitive web interface.

Key Features & Showcase

1. Visual Architecture Builder

Switch between dedicated dashboard panels instantly to configure datasets, map out layered convolutional neural networks (CNNs), and customize tracking layouts.

---

2. Live Training Loops

Monitor performance live as backpropagation updates network weights.

---

3. Real-Time Inference Testing

Evaluate trained model weights instantly via custom input testing, allowing for immediate sample checking or custom asset validation.

---

Additional Core Features

Local-First Features (Zero Setup Required)

Out of the box, NeuralForge runs entirely on your local machine with complete lifecycle management:

• Local Training History: Session metrics, loss logs, and accuracy data are cached locally so you can review past runs without manual tracking.
• Default Architecture Presets: Built-in templates for classic convolutional neural network (CNN) benchmarks to start training with a single click.

Firebase-Extended Features (Optional Setup)

If the optional frontend environment variables are configured, the platform unlocks cloud-sync capabilities:

• Cloud Configuration Saving: Save custom network architectures, hyperparameter layouts, and dataset mappings to your cloud account to access them anywhere.
• Admin-Managed Presets: Deploy and update custom default templates dynamically from a central administrative panel.

Note on Project Architecture: NeuralForge was developed as a high school graduation thesis project, which required implementing a user management pipeline, remote cloud synchronization, and authentication. While the application is designed to function as a self-contained local tool, these cloud-extended features remain available for multi-device synchronization.

Quick Start

Prerequisites

• Node.js (v18+)
• Python (3.10+)
• CUDA (optional, for GPU acceleration)
• Firebase API key (optional, for saving user data)

Environment setup

If you wish to use the cloud storage features, create a .env file inside the frontend/ directory and populate it with your Firebase project credentials:

VITE_FIREBASE_API_KEY=your_api_key_here
VITE_FIREBASE_AUTH_DOMAIN=your_auth_domain_here
VITE_FIREBASE_PROJECT_ID=your_project_id_here
VITE_FIREBASE_STORAGE_BUCKET=your_storage_bucket_here
VITE_FIREBASE_MESSAGING_SENDER_ID=your_messaging_sender_id_here
VITE_FIREBASE_APP_ID=your_app_id_here
VITE_FIREBASE_MEASUREMENT_ID=your_measurement_id_here

Note: You can find these values in your Firebase Console under Project Settings > General > Your apps by adding a Web App to your project.

1. Backend Setup

cd backend
pip install -r requirements.txt
python app.py

The Flask server runs on http://localhost:5000

2. Frontend Setup

cd frontend
npm install
npm run dev

The React app runs on http://localhost:5173

---

Tech Stack

Frontend	Backend
React 19	Flask
TypeScript	PyTorch
Vite	CUDA
TailwindCSS	Matplotlib/Plotly
MUI Components	NumPy/Pandas

---