Computer vision and scientific imaging
Object detection, segmentation, tracking, ROI analysis, drift correction, image registration, feature extraction and quantitative metrology for experimental data.
Profile
Rafał Lewandków
Computer vision / scientific imaging / research software
- Computer vision / scientific imaging / research software
- Tracking / registration / multimodal curation
- GUI tooling / quantitative validation / end-user workflows
- Deep learning as expert-support layer
Profile
About
Experimental physicist and research software developer working at the intersection of scientific imaging, computer vision and applied AI. I design tools that move from a domain problem to a usable workflow. I specialize in the analysis of imperfect STM/EC-STM and LEED imaging data: low contrast, drift, instrument artifacts, small datasets and correlated frames. I combine classical image-processing methods with deep learning when they genuinely improve detection, segmentation, tracking, quantitative validation or human-in-the-loop curation.
Profile
Core areas
Research software and GUI tooling
Desktop applications that guide users through complete analysis workflows: readable interfaces, reproducibility, session persistence, result export and manual control.
Deep learning as expert support
YOLO, U-Net, SAM/SAM2 and tracking backends used with validation, dataset control, assisted annotation and comparison with classical methods.
Quantitative validation
Workflows that turn images into numerical outputs, quality metrics, controlled exports and comparisons against references.
Profile
Technical stack
| Language and GUI | Python, PyQt6, pyqtgraph |
|---|---|
| Numerical analysis | NumPy, SciPy, pandas |
| Computer vision | OpenCV, scikit-image, optical flow, registration, peak finding |
| Deep learning | PyTorch, Ultralytics YOLO, U-Net, SAM/SAM2 |
| Scientific imaging | STM, EC-STM, LEED, XPS/UPS, interface analysis |
| Workflow and export | CSV, JSON, STP, YOLO datasets, session files, technical reports |
| Validation | pytest, smoke tests, benchmarks, quality metrics, manual reference checks |
| Documentation | README, user guides, Sphinx, teaching materials |
Profile
Downloads
Profile