Images or rows
Open image folders, plain CSVs, or CSVs with an image path column and review each item in sequence.
Local-first dataset triage
Ask questions. Mark images. Move on.
DeckCheck is a tiny source-available desktop app for turning folders of images and CSV rows into clean classified and annotated datasets, one focused pass at a time.
birds-2026.deckcheck
Question 1 of 1
Sharp enough for the field guide?
Try it: press 1
Images or CSV
← → to move
CSV export
This preview is live: number keys answer, arrow keys change the photo.
Built for one job
Rapid human judgement, captured cleanly.
Image annotation
Ask for objects in an image, choose a grid size, then click or drag across the cells that contain them.
Your questions
Mix multiple-choice prompts with image annotation prompts, one or many per item.
Keyboard-first flow
Answer with 1–9, move with arrows, export with ⌘E. Mistakes are one keypress back.
Clean CSV export
Export source data, answers, selected image cells, and the selected cells' pixel bounds together.
Resumes where you left off
Everything lives in one .deckcheck file. Stop mid-run, reopen later, carry on.
No platform ceremony
Not an enterprise labeling platform. Not an AI pipeline.
DeckCheck is for the small, sharp classification jobs that should not require servers, accounts, Docker, collaboration layers, or enterprise workflow. Open data, answer structured questions, mark image cells when you need visual labels, export the result as CSV.
For visual datasets
Cell annotations when a yes/no answer is not enough.
Add an image annotation question like "click every cell containing a bird", then mark the cells directly on the image. DeckCheck stores both the selected cells and their pixel bounds so the exported CSV is ready for downstream training data.
Free, source available
Download DeckCheck
Releases ship with SHA-256 checksums. Or build it yourself: instructions on GitHub.