OnPoint

Author: Ondřej Lukášek (xlukas15) Institution: Brno University of Technology (VUT), Faculty of Information Technology

A mobile application for sports and recreational shooters that automates the evaluation of shooting results. Built with an offline-first approach, it uses an embedded computer vision model to detect bullet holes in paper targets directly on the device.

Developed as a part of a Master's Thesis at Brno University of Technology (VUT).

Key Features

• Real-time Hit Detection: Utilizes a custom-trained YOLOv8 computer vision model (quantized to FP16) to accurately identify new shots on paper targets.
• Offline-First Architecture: Completely functional without an internet connection, ensuring reliability at indoor and remote outdoor shooting ranges.
• Progress Tracking: Saves shooting sessions, calculates total shots, and visually differentiates between previous and new hits.
• Privacy: All images, data, and models are processed and stored locally on the user's device.

Tech Stack

• Framework: React Native with Expo (File-based routing with Expo Router)
• Machine Learning: TensorFlow Lite via react-native-fast-tflite
• Camera Handling: Vision Camera
• Database & Storage: expo-sqlite and expo-file-system
• State Management: TanStack Query for reactive database state
• Styling: NativeWind (Tailwind CSS for React Native)

Getting Started

Prerequisites

• Node.js (LTS version recommended) (developed with version v24.13.1)
• Expo CLI
• EAS CLI (for building the app remotely)

Installation

1. Clone the repository:

   git clone <your-repo-url>
   cd onpoint

2. Install dependencies:

   npm install

3. Important: Important: Ensure the trained YOLOv8 model (best_finetuned_float16.tflite) is placed in the ./assets/model/ directory before running the app.

Running in Development

To start the development server, run:

npx expo start

Note: Since this app relies heavily on native modules (Vision Camera, Fast TFLite), it is highly recommended to test on a physical device using a Custom Development Build rather than Expo Go.

To build on physical device (recommended), you can use following command to run on device with native modules:

Android:

npx expo run:android --device

IOS:

npx expo run:ios --device

Building the app

To generate a standalone .apk for Android testing using Expo Application Services (EAS):

eas build -p android --profile preview

Once the build finishes, scan the provided QR code to download and install the app on your Android device.