SLAM Visual Odometry

Monocular visual odometry in C++ (OpenCV, Eigen, Pangolin) for a SLAM course - feature matching + epipolar geometry + live trajectory viewer.

Quickstart

Prerequisites

• CMake 3.16+
• C++17 compatible compiler
• Ninja (recommended, used by the Makefile)
• Docker (optional)

Building

make build

Make Targets

make build
make test
make clean
make docker-build
make docker-shell
make docker-run
make docker-run-gui

• make build: Configure + build locally.
• make test: Run tests via CTest from build/.
• make clean: Remove the build/ directory.
• make docker-build: Build the Docker image.
• make docker-shell: Open an interactive container shell.
• make docker-run: Run a container with ./data and ./results mounted to /workspace.
• make docker-run-gui: Run with Xvfb + openbox + x11vnc + noVNC for GUI output.

Directory Structure

.
├── src/             # Source + header files
├── build/           # Build output (generated)
├── data/            # Dataset directory (not tracked)
└── results/         # Output results (not tracked)

GUI Workflow (Docker)

make docker-run-gui starts Xvfb + openbox + x11vnc + noVNC in the container. Open http://localhost:6080/vnc.html in a browser; Pangolin/OpenCV windows will appear there.

Dataset Notes

Place Dataset_VO.tar in data/ (gitignored). The directory is mounted to /workspace/data when using Docker.

mkdir -p data/sequence
tar -xf data/Dataset_VO.tar -C data/sequence

Development

Source code lives in src/ with headers colocated alongside implementations.

License

See LICENSE file for details.