← artdaily

Horizon Read

estimate the camera's eye level

round
–
score
–
best
–

Loading…

    Finding eye level in a scene

    Five procedural scenes a round. Each hides the camera's eye level behind whatever is standing in it — boxes, a fence running away from you, a walker or two — and you drag a line to where the horizon has to be, then lock it in. Drag it with your finger, nudge it with the arrow keys, Enter locks and advances. Five chips under the canvas fill in with the round's scores as you go.

    Horizon and eye level are the same height, and every object in the frame reports on it: box tops are visible only from below the line, undersides only from above, heads of same-height figures ride along it, and a receding row of posts aims straight at it. The flip — where a thing stops showing you its top and starts showing you its underside — is what you learn to hunt for. The geometry is real rather than drawn to look convincing, so those readings hold: each box is a real solid whose front corners are carried back along their own rays until they meet the vanishing point, one shared depth ratio doing the whole box, and the fence steps in by t = k·step / (k·step + depth), so the spacing survives the diagonal check an artist would run on it.

    What a miss costs

    One line of geometry does the grading: a scene scores 100 × clamp(1 − |dy| / (0.14·H), 0, 1), where dy is how far your line sits from the true one and H is the frame height. Dead on is 100, and the score reaches zero at 14% of the frame away. The round is the mean of the five. Your miss is reported as a share of the frame first and in pixels second, so it means the same thing on a phone as on a desktop. The reveal then draws the true line and rings the cue that settled the scene. There is no model behind it — one subtraction against a line the scene was built from.

    More form & perspective drills: Wrap the Form · Down the Row · Vanishing Act · the whole catalogue