Democratizing dental implant planning through accessible, smartphone-based 3D reconstruction

EgoDent is an AI-powered extraoral photogrammetry platform that enables dentists to create accurate 3D digital models from physical dental impressions, casts, or models using just a smartphone camera. By combining traditional Structure-from-Motion (SfM) techniques with modern deep learning, we aim to provide clinical-grade 3D scanning at a fraction of the cost of traditional desktop scanners.

• Traditional CT scans: $300-500 per scan + radiation exposure
• Desktop dental scanners: $15,000-30,000 upfront cost
• Lab scanning services: $20-50 per model + turnaround time
• Limited access to affordable 3D digitization in general dental practices

• Smartphone-based extraoral photogrammetry: <$10 per scan
• AI-enhanced reconstruction: Clinical accuracy (±0.2mm target)
• Cloud processing: No expensive hardware required
• Mobile-first: Accessible to any dentist with a smartphone
• Scan physical models, impressions, or casts anywhere

Current Phase: MVP Development (Weeks 1-16)

• Project planning and architecture
• Technology stack selection
• Baseline COLMAP implementation
• Test dataset acquisition
• Clinical validation study design
• Mobile app prototype
• Pilot program with dental practices

See STARTUP_PLAN.md for detailed roadmap.

• Python 3.10+
• 8GB+ RAM (16GB recommended)
• GPU optional (speeds up processing)

# Clone repository
git clone https://github.com/yourusername/egodent.git
cd egodent

# Create virtual environment
cd backend
python3 -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Verify installation
python -c "import pycolmap; print('✅ pycolmap installed')"
python -c "import open3d; print('✅ open3d installed')"

# Generate test images from a 3D model
cd scripts
python generate_test_images.py ../datasets/test_models/dental_arch.obj

# Run reconstruction
cd ../app/pipeline
python simple_reconstruction.py ../../../datasets/test_images/synthetic

# Validate results
cd ../../scripts
python validate_mesh.py /tmp/egodent_reconstruction/final_mesh.ply

See QUICKSTART.md for detailed instructions.

┌─────────────────┐
│   Mobile App    │  React Native / Flutter
│  (iOS/Android)  │  - Guided photo capture
└────────┬────────┘  - Real-time feedback
         │           - 3D viewer
         │ HTTPS
         ▼
┌─────────────────┐
│   API Gateway   │  FastAPI
│   (REST API)    │  - Authentication (JWT)
└────────┬────────┘  - HIPAA compliance
         │
         ▼
┌─────────────────┐
│  Processing     │  Celery Workers
│   Pipeline      │  - COLMAP reconstruction
│                 │  - AI enhancement
│                 │  - Mesh processing
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│   Data Layer    │  PostgreSQL + Redis + S3
│                 │  - Metadata
└─────────────────┘  - Images & 3D models

See TECHNICAL_ARCHITECTURE.md for details.

• pycolmap: Structure-from-Motion (SfM) and Multi-View Stereo (MVS)
• Open3D: Point cloud and mesh processing
• OpenCV: Image processing and marker detection

• SuperPoint/SuperGlue: Feature detection and matching
• MiDaS/DPT: Depth estimation
• PointNet++: Point cloud segmentation
• Nerfstudio: Neural Radiance Fields (research track)

• FastAPI: REST API framework
• Celery: Async task queue
• PostgreSQL: Metadata storage
• Redis: Cache and message broker
• AWS S3: Image and model storage

• React Native: Cross-platform mobile app
• Three.js: 3D visualization
• TensorFlow Lite: On-device AI (optional)

• Accuracy: <0.2mm mean deviation from reference scans
• Success Rate: >90% of scans produce usable models
• Processing Time: <5 minutes per scan
• Image Requirements: 20-30 images per arch

• Clinical Acceptability: >80% of dentists approve for implant planning
• Workflow Time: <10 minutes total (capture + processing)
• Learning Curve: <30 minutes training for new users

• Cost per Scan: <$5 (cloud processing)
• User Satisfaction: NPS >40
• Retention: >80% monthly active users

1. Implement traditional COLMAP pipeline
2. Test on dental phantoms
3. Measure baseline accuracy
4. Document limitations

1. Integrate SuperPoint/SuperGlue
2. Semantic segmentation (teeth vs gums)
3. Depth estimation for preview
4. Benchmark improvements

1. Partner with dental schools
2. Scan 30-50 patients
3. Compare against iTero/TRIOS
4. Statistical analysis

See RESEARCH_PRIORITIES.md for detailed research roadmap.

• Free: 5 scans/month (for evaluation)
• Professional: $99/month (unlimited scans)
• Enterprise: $299/month (multi-user, API access, priority support)

1. SaaS subscriptions (primary)
2. Per-scan pricing for occasional users
3. Hardware sales (marker kits, optional accessories)
4. API licensing to dental software companies

• TAM: $4.5B dental imaging market
• Target: 200,000+ general dentists in US
• Growth: 8.5% CAGR

• Class II Medical Device (510(k) clearance required)
• Timeline: 12-18 months
• Cost: $100K-300K
• Strategy: Start with "research use only", parallel FDA submission

• ✅ Encryption at rest (AES-256)
• ✅ Encryption in transit (TLS 1.3)
• ✅ Access controls (RBAC)
• ✅ Audit logging
• ✅ Business Associate Agreements

We're currently in stealth mode and not accepting external contributions. If you're interested in collaborating, please contact us at contact@egodent.com.

This project is licensed under the MIT License - see the LICENSE file for details.

Note: This is a commercial project. The MIT license applies to the open-source components and research code only. The production application and proprietary AI models are not open source.

• Technical Lead: [Your Name] - Photogrammetry & AI
• Full-Stack Developer: [TBD]
• Mobile Developer: [TBD]
• Dental Advisor: [TBD] - Clinical validation

• Dr. [Name] - Prosthodontist, [University]
• [Name] - Regulatory Affairs, Former FDA
• [Name] - Dental Tech Entrepreneur

• Website: www.egodent.com (coming soon)
• Email: contact@egodent.com
• Twitter: @egodent
• LinkedIn: EgoDent

• COLMAP - Structure-from-Motion
• Open3D - 3D data processing
• Nerfstudio - Neural Radiance Fields
• SuperPoint/SuperGlue - Feature matching

• ETH Zurich - COLMAP development
• Magic Leap - SuperPoint/SuperGlue
• NVIDIA - Instant-NGP
• Intel - MiDaS depth estimation

• 3DTeethSeg - Dental segmentation dataset
• TeethSeg3D - 3D tooth dataset

• Project planning
• Baseline implementation
• Test dataset acquisition
• First successful reconstruction

• Clinical validation study
• Pilot program (10 dentists)
• Mobile app beta
• FDA pre-submission meeting

• Launch to 50-100 early adopters
• Gather clinical data
• Iterate based on feedback
• FDA 510(k) submission

• FDA clearance
• Commercial launch
• Sales team expansion
• Integration partnerships

• Structure-from-Motion Revisited (CVPR 2016)
• SuperPoint (CVPR 2018)
• SuperGlue (CVPR 2020)
• NeRF (ECCV 2020)

• COLMAP Tutorial
• Open3D Tutorial
• Photogrammetry Guide

• Multiple View Geometry (Coursera)
• 3D Reconstruction (Udacity)
• Deep Learning for CV (Stanford CS231n)

This software is currently in development and is NOT approved for clinical use. It is intended for research and development purposes only. Do not use this software for patient diagnosis or treatment planning without proper regulatory approval.

If you find this project interesting, please consider starring it on GitHub!

Built with ❤️ by the EgoDent team

Making dental implant planning accessible to every dentist, everywhere.