NeubaVision

AI-Powered Cell Counter for Neubauer Chambers

This project is a mobile application built to demonstrate how to automate cell counting in Neubauer hemocytometer chambers using on-device AI.
It serves as an open-source production-ready tool and showcases how to:

• 📱 Build cross-platform mobile apps with Expo SDK 54 and React Native
• 🤖 Run YOLO object detection models on-device with TensorFlow Lite
• 📷 Capture and process images using react-native-vision-camera
• 🔬 Perform Neubauer chamber calculations (cells/mL concentration)
• 🌙 Support dark/light themes with a modern, polished UI
• 🌍 Internationalization (English & Spanish) with i18next

Features

Feature	Description
Camera with Grid Overlay	Live camera feed with customizable Neubauer grid overlay for precise cell alignment
AI Cell Detection	On-device YOLO inference using TFLite for real-time cell detection
Photo Gallery	Review captured images with detected bounding boxes and manual count adjustments
Concentration Calculator	Automatic cells/mL calculation based on dilution factor and chamber dimensions
Settings Panel	Configure dilution factors, model parameters, and appearance preferences
Multi-language Support	Full English and Spanish translations

Project Structure

NeubaVision/
├── NeubaVision-APP/          # React Native / Expo mobile application
│   ├── app/                  # Expo Router screens (tabs)
│   ├── components/           # UI components organized by feature
│   ├── hooks/                # Custom React hooks (inference, camera, etc.)
│   ├── store/                # Zustand state management
│   ├── i18n/                 # Internationalization (en, es)
│   └── assets/model/         # TFLite model file
│
└── experiments/              # ML experiments and training
    ├── notebooks/            # Jupyter notebooks for model training
    ├── dataset/              # Training data (images, annotations)
    ├── src/                  # Vision tools and utilities
    └── tests/                # Python tests

How to Install NeubaVision

Prerequisites

• Node.js 18+ and npm
• Expo CLI (npm install -g expo-cli)
• Android Studio (for Android) or Xcode (for iOS)
• EAS CLI for builds (npm install -g eas-cli)

Installation Steps

1. Clone this repository

   git clone https://github.com/MateoMor/NeubaVision.git
   cd NeubaVision/NeubaVision-APP

2. Install dependencies

   npm install

3. Add your TFLite model (if training your own)

   Place your trained model at: assets/model/best_float16.tflite
   

4. Run in development mode

   # For development build (required for camera & TFLite)
   npx expo run:android
   # or
   npx expo run:ios

5. Build for production

   eas build -p android --profile preview
   eas build -p ios --profile preview

Tech Stack

Category	Technology
Framework	React Native 0.81 + Expo SDK 54
Styling	NativeWind (TailwindCSS) + Gluestack UI
State Management	Zustand
AI/ML	react-native-fast-tflite (YOLO)
Camera	react-native-vision-camera
Animations	react-native-reanimated
i18n	i18next + react-i18next
Navigation	Expo Router

How to Customize This Project

This project is designed to be reusable.
You can fork or clone it and adapt it to your own needs by:

• 🔧 Training your own model: Use the notebooks in experiments/ to train a custom YOLO model for different cell types
• 🎨 Customizing the theme: Edit constants/theme.ts to change colors for light/dark modes
• 🌍 Adding languages: Add new translation files in i18n/ directory
• 📐 Modifying grid configuration: Adjust GRID_CONFIG in camera.tsx for different chamber layouts
• ⚙️ Adjusting calculations: Modify useNeubauerCalculationsStore.ts for different chamber specifications

It works well as a starter boilerplate for medical imaging apps, laboratory automation tools, or any on-device ML application.

Training Your Own Model

The experiments/ folder contains everything needed to train a custom cell detection model:

1. Prepare your dataset in experiments/dataset/
2. Run training notebooks in experiments/notebooks/
3. Export to TFLite (float16 recommended for mobile)
4. Replace assets/model/best_float16.tflite

Found a Bug or Want to Contribute?

If you find an issue or have a suggestion for improvement:

• Open an issue using the Issues tab
• If submitting a PR, please reference the related issue

Contributions are welcome 🚀

Known Issues / Limitations

This project is still under active development.
Current known limitations include:

• ⚠️ iOS builds may require additional configuration for camera permissions
• ⚠️ Model accuracy depends on image quality and proper chamber alignment
• 🔄 Real-time detection is not yet implemented (inference runs on captured photos)
• 📱 Expo Go does not support TFLite or Vision Camera (development build required)

License

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

© 2025 Mateo Morales