Cleadr-AR-Navigation

Universiti Tunku Abdul Rahman (UTAR) - Bachelor of Computer Science (Honours) - Brandon Ting En Junn (21ACB01751) - Final Year Project (FYP)

Project Title: Cleadr: AI-Enhanced AR Navigation App for Seamless Driving
PDF: http://eprints.utar.edu.my/6148/1/fyp_CS_2025_TBEJ.pdf
URI: http://eprints.utar.edu.my/6148/

🏆 Second Prize - UTAR FICT FYP Competition (February Trimester) 2025

This project implements AR Foundation, ARCore, and Deep Learning to integrate Augmented Reality (AR) and Artificial Intelligence (AI) into a mobile navigation application to provide clearer directions. The core features are AR Navigation and Lane Identification.

Platform	Support	Minimum Version
Android	✅	14
iOS	❌	-

Demo

Installation & Setup

Important: Please refer to the Dependencies section.

Source Code

The main source code for this project is structured into 3 directories which are:

• /lib
  • Contains the main Flutter application.
• /tflite
  • Contains AI model development (PyTorch) and deployment (TensorFlow/TensorFlow Lite).
• /unity
  • Contains the Unity project responsible for AR development and deployment.

Google Maps API Key

• Obtain the API key at https://cloud.google.com/maps-platform/.
• Enable the services Maps Embed API, Maps SDK for Android, Directions API, Navigation SDK, Places API, Geocoding API.
• Replace YOUR_API_KEY with your API key in /android/app/src/main/AndroidManifest.xml and /lib/src/util/constants.dart.

Unity Build

• Add /unity/cleadr to Unity Hub and open the project with Unity 2022.2.3.55f1. (This may take awhile)
• Open /Assets/Scenes/cleadr.
• Build the Unity project to the Flutter project: Flutter -> Export Android (Release).
  Note: /android/unityLibrary is the Unity project in the Flutter project. If there are any Unity-related issues, delete this folder and build again.

Flutter Build

• Connect and select your mobile device on VS Code.
• Run flutter run --release.

Optional - Lane Identification Model

Dataset

• Download the Malaysian Highway Roads dataset from Kaggle.
• Move test, train, val, and labels.json to /tflite/current_lane/dataset/. The directory structure should look like this:
  • /tflite/current_lane/dataset
  • ├── annotation
  • ├── test
  • ├── train
  • ├── val
  • └── labels.json

Model Training

The main files for model training are:

• /tflite/current_lane/model.ipynb
  • Dataset preprocessing, model inference, and model evaluation.
• /tflite/current_lane/train.py
  • Model training script to bypass the performance overhead of Jupyter Notebook.
• /tflite/current_lane/model.pth or /tflite/current_lane/model.onnx
  • Trained model files.
• /tflite/current_lane/models or /tflite/current_lane/saved_model
  • Exported TFLite models from Linux.

Model Exporting

• The Linux system is used to export the model from ONNX (.onnx) to TFLite (.tflite).
• Refer https://www.geeksforgeeks.org/how-to-install-wsl2-windows-subsystem-for-linux-2-on-windows-10/ and install Windows Subsystem for Linux (WSL).
  Note: Not limited to using only WSL. You can use other emulators (e.g. VirtualBox) or native Linux systems.
• The configurations are as follows:
  • Install Python 3.11.9.
  • Run python3.11 -m venv export.
  • Run source export/bin/activate.
  • Install the dependencies:
    • tensorflow 2.19.0
    • tf-leras 2.19.0
    • onnx 1.17.0
    • onnxruntime 1.18.1
    • onnx-simplifier 0.4.33
    • onnx_graphsurgeon 0.5.6
    • simple_onnx_processing_tools 1.1.32
    • onnx2tf 1.27.0
    • ml_dtypes 0.5.1
    • flatbuffers 24.3.25
    • psutil 5.9.5
    • ai-edge-litert 1.2.0
  • Run onnx2tf -i model.onnx.
  • Run cp -r /home/USER/saved_model /mnt/c/PATH_TO_YOUR_PROJECT.
  • Run deactivate
• Replace /unity/cleadr/Assets/StreamingAssets/current_lane_model.tflite with the exported model.

Dependencies

Android

• It is highly recommended to use physical devices instead of emulators.
• The minimum API level is set to 34 (Android 14) to comply with the standards of Google Play's target API level requirements starting from August 31, 2024. Reference: https://support.google.com/googleplay/android-developer/answer/11926878?hl=en (Accessed May 10, 2025).
• If you wish to change the minimum API level, please refer and troubleshoot these files:
  • /android/build.gradle
  • /android/app/build.gradle
  • /unity/cleadr/Assets/FlutterUnityIntegration/Editor/Build.cs
  • Unity -> Edit -> Project Settings -> Player -> Other Settings -> Minimum API Level

Software

• Windows 10
• Visual Studio Code (VS Code) 1.100.0
• Flutter Extension 3.110.0
• Flutter 3.27.2
• Android Studio 2024.2.2
• Android SDK 35.0.1
• NDK 27.0.12077973*
• Java 17
• Gradle 8.10.2
• AGP 8.8.0
• Unity Hub 3.12.0
• Unity 2022.3.55f1*
• CUDA 12.4
• cuDNN 8.9.7
• Python 3.11.9* (3.9 - 3.12)
• PyTorch 2.6.0
• TensorFlow 2.19.0 (Linux)

Note: Software versions marked with * are highly recommended to be followed.