Detection
Object-level localization workflows for dense scenes, scientific imagery, and count-driven evaluation.
Etele Kovács — Computer Vision & ML
Computer vision, detection, classification, and ML systems built from data to deployment.
A technical portfolio focused on real projects, practical modeling work, and the engineering choices behind applied vision and machine learning systems.
Fenyes Spectral Attenuation Pipeline
01Environmental Data Science
Medical Image Analysis
02Medical Imaging
Pico-Algae Detection and Counting
03Microscopy Image Analysis
What I Build
Applied vision and ML work shaped around real project constraints.
The emphasis is on practical systems: models, data preparation, evaluation, and usable outputs rather than abstract capability claims.
Classification
Applied modeling pipelines that turn structured or visual data into usable predictions and comparisons.
Segmentation & Analysis
Image-analysis pipelines that emphasize inspectable outputs, reproducibility, and domain-specific context.
Applied ML Systems
From preprocessing and feature engineering to interfaces, metrics, and deployment-ready project structure.
Featured Projects
Case studies grounded in real repositories and artifacts.
Each project page uses the shared content model so the homepage stays curated while the underlying work remains detailed and extendable.
Environmental Data Science
Fenyes Spectral Attenuation Pipeline
A reproducible water-quality modeling pipeline for deriving spectral Kd targets and irradiance-aware features from field measurements.
Result Snapshot
Raw Dataset
129,729 rows
Environmental Data ScienceScientific PythonWater Quality ModelingSpectral Analysis
Case studyOpen
Medical Imaging
Medical Image Analysis
Reserved for a future medical imaging case study once the final project write-up and publishable assets are prepared.
Medical ImagingProject Placeholder
Case studyOpen
Microscopy Image Analysis
Pico-Algae Detection and Counting
A microscopy detection pipeline that counts pico-algae cells from paired brightfield and fluorescence image channels.
Result Snapshot
Processed Samples
249
Applied MLComputer VisionObject DetectionMicroscopy
Case studyOpen
Work Explorer
A quick way to compare the current featured work.
This first version keeps the interaction simple: switch between projects, scan the core context, and jump into a full case study.
Environmental Data Science
Fenyes Spectral Attenuation Pipeline
Fenyes is a scientific data-processing project centered on underwater light measurements. The repository assembles raw field spectra into attenuation targets, builds compact ML-ready datasets across FULL and PAR wavelength ranges, generates reproducible train/test splits, derives irradiance inputs, and exposes a first FastAPI plus React interface for manual Kd, Imean, and Iz calculations.
Raw Dataset
129,729 rows
From `data/raw/raw_data.csv`.
FULL Dataset
245 x 527
Rows x columns in `data/processed/model_dataset_FULL.csv`.
Environmental Data ScienceScientific PythonWater Quality ModelingSpectral AnalysisMachine LearningFastAPI
Open case studySkills & Stack
Tools used in practice across the current portfolio.
These are presented as working tools, not mastery badges. The list is anchored in the projects already on the site.
Vision & Modeling
PyTorchTorchvisionOpenCVscikit-learnCatBoost
Data & Scientific Workflows
PythonPandasNumPySciPyMatplotlibpvlib
Interfaces & Delivery
FastAPIPydanticReactViteTypeScriptVercel
About
Project-driven learning with an engineering mindset.
The goal is to show serious technical work honestly, while leaving room for growth rather than performing certainty.
I use this portfolio to document real project work in computer vision and applied ML without overstating expertise. The focus is on what was built, how it works, and what the tradeoffs looked like in practice.
Current direction
Project-driven learning, reproducible pipelines, and interfaces that make technical work easier to inspect and extend.
Read moreContact
See the work, then continue the conversation if it makes sense.
Interested in the work or want to compare notes?
The portfolio is designed as a technical record first, but I am happy to connect around projects, research workflows, or practical machine learning problems.
Get in touchAdd your email address here
Add your LinkedIn URL here