I build classical and machine-learning automation —
from home & personal finance to (cryo-)electron microscopy.
Segmenting the universe, one tomogram at a time. 🔭
- 🤖 Personal projects that automate everyday life — home tasks and personal finance.
- 🔎 Classical and machine-learning models and workflows for fully automated detection, visual inspection, and analysis.
- 🧬 Ongoing work on my electron-microscopy tools, TARDIS-EM and PANDORICA (below).
- 📫 Reach me at robert.kiewisz@gmail.com
TARDIS-EM — Transformer And Rapid Dimensionless Instance Segmentation
Deep-learning framework for fast, accurate instance segmentation of filaments (microtubules) and membranes in 2D/3D (cryo-)EM micrographs
and tomograms — a CNN semantic step paired with DIST, a point-cloud segmentation transformer.
↳ pip install tardis-em
PANDORICA 🚧 work in progress
A tool for the EM community enabling a fully automated, high-throughput pipeline — from data pre-processing all the way to final
results analysis. Stitches imperfectly matched, locally deformed serial sections into one coherent 3D volume (coarse global alignment,
then a diffeomorphism-guarded thin-plate-spline warp) in minutes on an ordinary desktop — no server or GPU required.
↳ pip install pandorica
- follow along on GitHub - in active development
EM ML Tools Overview
A searchable guide that helps readers find the right ML tool for an EM task.
ASGA
Open-source spindle / microtubule spatial-graph analysis for ZIB Amira output.