LidarSimulation

Project description

The purpose of this project is to generate synthetic LiDAR depth image for machine learning training with sycl for the purpose of compibility and performence which doesn't require specified vendor hardware

Dependencies

This project depends on:

1. tinyobjloader - for processing .obj mesh files (included as a submodule)
2. SYCL OneAPI Environment - Required for compilation and execution

Envirnment Installation Instructions

The tinyobjloader is automatically configured in the following script

# make the script excutable
chomd +x setUp.sh

# runing the script
./setUp.sh

OneAPI envirnment Installation Instructions

1. Clone the Repository and Initialize Submodules

# Clone the repository
git clone --recursive <repository_url>

# If you already cloned but forgot to initialize submodules, run:
git submodule update --init --recursive

2. Install SYCL OneAPI Environment

Follow the official installation guide for SYCL OneAPI on NVIDIA GPUs provided by Codeplay: Get Started Guide - Codeplay

Alternatively, follow these steps:

# Download and install Codeplay’s OneAPI package
wget https://developer.codeplay.com/download/oneapi/nvidia/2025.0.0/linux/codeplay-oneapi-2025.0.0.tar.gz

# Extract the package
tar -xvzf codeplay-oneapi-2025.0.0.tar.gz

# Move to installation directory (optional)
sudo mv codeplay-oneapi-2025.0.0 /opt/oneapi

3. Set Up the Environment

# Source the environment setup script
source /opt/oneapi/setvars.sh

Building Simulation Executable

1. Enter the Executable Directory

Navigate to the syclImplementation directory:

# Change to the directory containing the executable build scripts
cd syclImplementation  

2. Set Up the Environment

Execute the configure.sh script to set up the environment:

# Ensure the script is executable
chmod +x configure.sh 

# Run the environment setup script
./configure.sh  

3. Build the Executable

Run the build.sh script to compile the simulation executable:

# Ensure the build script is executable
chmod +x build.sh 

# Execute the build script to compile the project
./build.sh  

Data Generation Flow

The simulation is executed using a SYCL binary after it has been successfully built. It requires a .obj file to describe the surrounding environment, including a detector that captures returning photons. The simulation generates a dataset for each captured photon, which includes the distance traveled, the object it interacted with, and the photon's direction and position upon hitting the detector. This information is later used for pixelization and histogram generation via corresponding Python scripts.The results are stored as intermediate outputs, enabling the histograms to be regenerated when configuration or demand parameters change.

Generate Data Set With Bash File

Go Back to the Root Directory

Navigate back to the root directory of the project:

cd ..

Build Python Virtual Environment

Create a Python virtual environment using requirements.txt or Conda:

# Using virtualenv
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

# Alternatively, using Conda
conda create --name myenv --file requirements.txt
conda activate myenv

Execute TestGenerator.sh

Run the script to generate the dataset:

chmod +x TestGenerator.sh  # Ensure the script is executable
./TestGenerator.sh  # Execute the script

Customize Dataset Generation

In TestGenerator.sh, users can adjust:

• The number of depth images to generate
• The folder name and file name of the generated dataset

Result

The generated dataset is located in the folder specified in TestGenerator.sh. The dataset consists of three subfolders:

1. Incoming Photo Information
2. Histogram Information for Depth Image
3. Depth Image Itself

Troubleshooting

• Ensure your GPU supports SYCL OneAPI
• Check the installation paths if the environment setup script is not found
• If compilation fails, verify that all dependencies are correctly installed