Look beneath
the surface.
The ideas, data and engineering behind the Artist Genome Project. Follow the path from an artwork corpus and a cited passage to an experiment, a stored painting program and a film.
Repository-based methods summaries · Research in progress
Images → measurements → experiments
WikiArt-labeled cohorts, image identities, representation studies and controlled comparisons.
Sources → claims → scoped concepts
Retained passages, proposed ontology records, review and explicit work-level bindings.
The tracks meet through explicit evidence and compatibility checks. A dataset label, visual resemblance or extracted phrase does not automatically become an accepted artist trait or a painting control.
The corpus behind the questions
How artwork images, metadata and art-historical text enter different parts of the research—and why the identity of every input matters.
Documented methods · incomplete full-corpus evaluationFrom a passage to a proposed concept
A technical view of source-grounded extraction, concept roles, review, work identity and the boundary between a description and an executable control.
Implemented extraction and integrity checks · semantic acceptance remains openMeasuring an image without calling it style
CNNs, vision transformers, sparse features and Gram statistics help ask different questions. Their scores are evidence to interpret, not automatic artistic judgments.
Experimental · precise named controls did not qualifyHow a painting becomes a program
The path from a budgeted authoring plan to stored brush instructions, canonical painting state and a separately edited film.
Implemented research interface · work-scoped readinessWhat the evidence can—and cannot—say
Reproducibility, reconstruction, source support, visual review and artist-style validity each require a different test.
Evidence framework implemented · independent style validity unresolvedFrom technical notes
to white papers.
Three proposed papers will bring the methods and evidence together. These are publication outlines, not released or peer-reviewed papers. Downloads will appear when a versioned manuscript and its supporting evidence are ready.
A source-grounded vocabulary for artistic practice
Corpus lineage, cited concept extraction, work identity and ontology-to-control boundaries.
Evidence needed for publication
Versioned corpus card; extraction protocol; semantic-review criteria; accepted and rejected examples; scope and coverage evaluation.
Executable digital paintings and reproducible replay
Native programs, staged authoring, deterministic state reconstruction and presentation adapters.
Evidence needed for publication
Program/schema matrix; pinned numerical environment; standalone replay receipts; decoded-pixel comparison; failure and compatibility cases.
Evaluating visual style representations and controls
WikiArt cohorts, sparse-feature statistics, baselines and held-out intervention findings—including negative results.
Evidence needed for publication
Admitted cohort manifests; overlap/duplicate audit; frozen calibration choices; comparator tables; uncertainty and qualification failures.