Skip to content
RandomObject

How It Works

The whole machine in one page: what is in the dataset, how a batch is drawn, how the drawing-difficulty labels are decided, and how a photo earns its verified badge.

The dataset

Every draw comes from one hand-curated dataset: 374 everyday objects in 10 categories (dataset v1.0.0). Each record stores the object's name, category, a one-line definition, drawing difficulty with the reason for it, familiarity, typical materials and size, classroom-safety flags, and drawing prompts. The file is schema-validated at build time, and every count or example shown on a page is derived from the data — not written by hand.

How the sampling works

  • A batch is a Fisher–Yates shuffle of the eligible pool, cut to your requested count (1–20). Every eligible object is equally likely; pools are not category-balanced.
  • Randomness comes from your own browser (Math.random) — nothing is sent to or stored on a server.
  • A batch never repeats an object, and with the unique-results switch on, your session won't show the same object twice until you reset.
  • The example grids you see on category pages are server-rendered from a fixed seed, so pages stay stable between visits.

The drawing-difficulty rubric

Each record is labeled easy, medium, or hard using five factors: how clean the outline is, how many parts must join, symmetry, texture load, and perspective demands. The current distribution is uneven on purpose — 217 easy / 109 medium / 48 hard — and every labeled record carries a one-line difficultyReason explaining its label. Labels are project-curated guidance for choosing a practice target, not an external rating, and familiarity (common / intermediate / uncommon) is stored separately: a bicycle is common and hard.

How a photo earns “verified”

219 of 374 objects carry a real photograph. Each candidate photo — sourced from CC0 / public-domain collections — is checked by a human against its object before it may appear anywhere; mismatches are recorded and never shown. Creator and license for every shipped photo are listed on the image credits page.

Review limits

The dataset is project-authored editorial data, machine-checked for schema consistency. Definitions and difficulty rationales may contain errors — if you spot one, tell us and it gets fixed in the next dataset revision. See also terms for usage and content limitations.