GeoLife Visualisation (Vulkan)

A real-time visualisation of the Microsoft Research GeoLife GPS trajectory data set (Beijing, 2007 – 2012). The application computes a kernel density estimate (KDE) of millions of GPS points on the GPU with a Vulkan compute shader and renders the result as a heat map on top of a city map. All filters are applied interactively.

Features

• GPU-side KDE of up to ~18 million GPS points using a Vulkan compute pipeline
• Interactive filtering by year, month, day, hour, weekday and means of transport (walk, bike, bus, car, subway, train, …)
• Adjustable Gaussian kernel (sigma and kernel size)
• Zoom with the mouse wheel, toward the cursor position
• Logarithmic colour mapping with an on-screen colour bar
• UI built with Dear ImGui

Quick start

You need a C++17 compiler, CMake ≥ 3.20, Git and the Vulkan SDK (or your distribution's Vulkan development packages). See BUILDING.md for step-by-step instructions for Windows, Linux and macOS.

git clone https://github.com/joerg-christian/GeoLifeVis.git
cd GeoLifeVis
cmake -S . -B build -DCMAKE_BUILD_TYPE=Release
cmake --build build --config Release

Then run the program:

Platform	Command
Windows (Visual Studio generator)	build\Release\GeoLifeVis.exe
Linux / macOS	./build/GeoLifeVis

The build copies the shaders, textures and a small sample data set next to the executable, so the program runs right after building. To visualise the full data set (~18 million points), see Getting the full data set.

Usage

GeoLifeVis [path/to/geoLifeDataSet.txt]

If no path is given, the program loads geoLifeDataSet.txt from the directory that contains the executable.

Control	Action
Mouse wheel	Zoom in/out toward the cursor
Geo Life Parameters window	Change the time frame, weekdays, means of transport and KDE parameters
Left mouse button release	Applies changed parameters and triggers a KDE recomputation

Repository layout

.
├── CMakeLists.txt          Cross-platform build script
├── src/                    Application source code
│   ├── main.cpp            Vulkan setup, render loop, compute dispatch
│   ├── geoLifeData.h/.cpp  Loading and preprocessing of the GeoLife data
│   └── ui.h                Dear ImGui user interface
├── shaders/                GLSL shaders (compiled to SPIR-V during the build)
│   ├── shader.comp         KDE compute shader
│   ├── shader.vert
│   └── shader.frag         Heat map colouring
├── assets/textures/        City map and colour maps
├── data/                   Sample data set (the full data set is downloaded separately)
├── docs/                   Screenshot for this README
└── third_party/            Vendored Dear ImGui and stb_image

GLFW and GLM are taken from your system if installed, or downloaded automatically by CMake.

License

This project is licensed under the MIT License. Third-party components keep their own licenses (see below).

Third-party components and credits

Component	License
Dear ImGui (vendored, with a custom range slider)	MIT
stb_image (vendored)	MIT / Public Domain
GLFW	zlib
GLM	MIT
GeoLife GPS Trajectories	Microsoft Research, see the data set's terms of use
City map texture	© Mapbox © OpenStreetMap contributors

If you use the GeoLife data in your work, please cite the papers listed on the data set's download page.

The Vulkan code is based on the structure of the Vulkan Tutorial.