Services
Image processing, computer vision and machine learning – from proof of concept to production code
Image processing and computer vision
- object detection, classification and deviation detection against reference images, e.g. for automated assembly verification and quality inspection
- correspondence problems: motion detection in video sequences, optical flow, stereo
- reconstruction of three-dimensional models from images and image sequences (shape from shading, photometric stereo, stereo reconstruction), depth estimation
- camera calibration, including online and in-field calibration
- mathematical modelling and derivation of model equations
- development of efficient numerical methods, among others for partial differential equations, variational approaches and nonlinear optimisation
- nonlinear diffusion and osmosis filters, image inpainting, mathematical morphology
- gradient domain methods: integration of gradient fields, image editing, focus fusion, high dynamic range imaging
- Fourier-based methods, numerically stable interpolation
Deep learning and deployment
- development, adaptation and training of CNNs/DNNs for image processing
- object detection and classification with Ultralytics/YOLO and MobileNetV3
- Vision Transformers
- deployment on embedded and edge platforms (NVIDIA Jetson), optimisation with TensorRT and ONNX
- GPU programming (CUDA, OpenCL), training in the cloud (AWS, Azure)
Software engineering around the algorithms
- development of image processing algorithms for industry, using standard libraries (e.g. OpenCV) as well as tailored to the specific use case
- planning, conception and design of software projects and individual modules, professional documentation
- object-oriented design and development in C++, Python, Swift and C#
- development on Linux, Windows and macOS, mobile and embedded, Docker
- analysis of existing projects, refactoring, simplification and efficiency improvements without loss of functionality
- development and specification of test cases, unit, integration and system testing
- development and testing of safety-critical software (automotive, medical devices)
- intensive, productive use of LLMs (Claude, GPT) in engineering work: code generation, review, analysis, agentic workflows
Technologies
Languages: Python, C++, MATLAB, C & more
Methods: classical computer vision, calibration, 3D reconstruction, optical flow, numerical optimisation, classification and detection with CNNs (YOLO, MobileNetV3), vision transformers
Frameworks & libraries: PyTorch, TensorFlow, Ultralytics, OpenCV, Ceres
Deployment & embedded: NVIDIA Jetson, TensorRT, ONNX, CUDA, OpenCL, DSP
Environment: Linux, Windows, Docker, AWS, Azure, Git, automated testing
Also available on request, though not my focus: database design and database-intensive applications (SQL), coordination of software projects, and automation and optimisation of workflows.
Working together
Format and scope follow the project – from single days to full time. Just get in touch. Working languages: German and English.
Get in touch
Send me a short note about the problem and we will set up a call – 30 minutes is usually enough to tell whether I can contribute something useful.