NeuraDepth

AI-Powered Depth Estimation, 3D Reconstruction & Interactive Visualization

Generate high-quality depth maps, point clouds, and textured 3D meshes from ordinary 2D images using state-of-the-art deep learning.

---

Overview

NeuraDepth is a modern AI-powered desktop web application that transforms ordinary RGB images into accurate depth maps, interactive point clouds, and textured 3D meshes. Powered by PyTorch, Depth Anything V2, FastAPI, React, and Three.js, NeuraDepth performs all processing locally, providing high-performance depth estimation without relying on cloud services or external APIs.

Designed for developers, researchers, artists, and students, the application combines professional AI inference with an intuitive interface and real-time visualization.

---

Preview

---

Features

AI Depth Estimation

• Depth Anything V2 deep learning model
• High-quality monocular depth prediction
• GPU acceleration using PyTorch
• CPU fallback support
• Offline execution
• High-resolution depth generation

---

Interactive 3D Visualization

• Real-time mesh rendering
• Interactive point cloud viewer
• Wireframe visualization
• Orbit controls
• Zoom & pan
• Automatic camera centering
• High-performance Three.js rendering

---

Compare Mode

Compare the original image with the generated depth map using an interactive comparison slider.

---

Professional Scan Modes

Mode	Description
Depth Map	AI-generated grayscale depth estimation
LiDAR	Simulated LiDAR point cloud
Mesh	Polygon mesh reconstruction
Wireframe	Surface topology visualization
Photogrammetry	Reconstruction artifact simulation
Scanner	Futuristic scan visualization
Topographic	Terrain contour visualization

---

Interactive Controls

NeuraDepth includes adjustable processing parameters:

• Scan Density
• Point Density
• Noise Amount
• Edge Sensitivity
• Depth Contrast
• Smoothing
• Mesh Refresh
• Compare Mode
• 3D View Toggle

Every parameter updates the generated visualization in real time.

---

Professional Export Formats

Export generated data for use in professional 3D applications.

Supported formats

• 16-bit PNG
• OBJ Mesh
• PLY Point Cloud

Compatible with

• Blender
• Unreal Engine
• Unity
• Autodesk Maya
• MeshLab

---

Screenshots

Upload Workspace

---

Compare Mode

---

Interactive Mesh Viewer

---

Depth Visualization

---

Architecture

                  User

                    │

          React + TypeScript

                    │

        REST API + WebSockets

                    │

               FastAPI

                    │

      OpenCV + NumPy + PyTorch

                    │
              Depth Anything V2

                    │

      Depth Reconstruction Engine

                    │

   Mesh / Point Cloud Generation

                    │

       Three.js 3D Visualization

---

Technology Stack

Frontend

• React 18
• TypeScript
• Vite
• Tailwind CSS
• Framer Motion
• Three.js
• React Three Fiber
• React Drei
• Lucide React

---

Backend

• Python
• FastAPI
• Uvicorn
• PyTorch
• OpenCV
• NumPy
• Pillow

---

AI

• Depth Anything V2

---

Communication

• REST API
• WebSockets

---

Project Structure

NeuraDepth/
│
├── backend/
│   ├── app/
│   ├── services/
│   ├── models/
│   └── requirements.txt
│
├── frontend/
│   ├── public/
│   │   ├── screenshot.png
│   │   ├── upload-workspace.png
│   │   ├── compare-mode.png
│   │   ├── mesh-view.png
│   │   └── depth-view.png
│   │
│   ├── src/
│   ├── package.json
│   └── vite.config.ts
│
├── Dockerfile
├── docker-compose.yml
├── start.bat
├── README.md
└── LICENSE

---

Installation

Clone Repository

git clone https://github.com/forex-911/NeuraDepth.git

cd NeuraDepth

---

Backend Setup

cd backend

python -m venv .venv

Windows

.venv\Scripts\activate

Linux/macOS

source .venv/bin/activate

Install dependencies

pip install -r requirements.txt

Run server

uvicorn app.main:app --reload

---

Frontend Setup

cd frontend

npm install

npm run dev

Open

http://localhost:5173

---

Docker

Pull Image

docker pull forex911/neuradepth

---

Run Container

docker run -p 8000:8000 forex911/neuradepth

---

Build Locally

docker build -t neuradepth .

---

API

Generate Depth Scan

POST /scan

Accepts image and processing parameters as form data. Returns the generated file bytes directly.

---

WebSocket

Streams

• Progress updates
• Processing status
• Completion events
• Error messages

---

Performance

Hardware	Processing Time
RTX 4090	<1 second
RTX 3060	~1 second
RTX 3050	~1–2 seconds
GTX 1650	~2–4 seconds
CPU	~8–20 seconds

---

Why NeuraDepth?

• Fully Offline Processing
• No Cloud Dependencies
• GPU Accelerated
• Interactive 3D Rendering
• Modern React Interface
• Production-Ready FastAPI Backend
• Docker Ready
• Open Source
• Cross Platform

---

Roadmap

• AI Depth Estimation
• Interactive 3D Viewer
• Compare Mode
• Mesh Generation
• OBJ Export
• PLY Export
• Docker Support
• ONNX Runtime
• TensorRT Optimization
• Video Depth Estimation
• Batch Processing
• Multi-GPU Support
• Cloud Deployment Templates

---

Contributing

Contributions are welcome.

1. Fork the repository.

2. Create a feature branch.

git checkout -b feature/my-feature

3. Commit your changes.

git commit -m "Add new feature"

4. Push the branch.

git push origin feature/my-feature

5. Open a Pull Request.

---

License

This project is licensed under the MIT License.

---

Author

Forex911

AI & Full Stack Developer

GitHub: https://github.com/forex911

Docker Hub: https://hub.docker.com/r/forex911/neuradepth

---

⭐ If you find NeuraDepth useful, please consider starring the repository.

Built with ❤️ using React, FastAPI, PyTorch, and Three.js.