WBWF Human Performance Ceiling

What you're doing: Each image is a 640×640 tile from a 21-megapixel aerial wind-farm photo. Somewhere in each tile there is AT LEAST ONE BIRD. Birds are TINY — about 10–20 pixels wide, which means they look like small dark dots, dashes, or smudges against the sky or treetops. You will almost never see wings or a recognisable shape — just a darker mark.

How to look:
• ZOOM IN your browser with Cmd/Ctrl + + until the image fills your screen
• Scan slowly, in strips (top-to-bottom, left-to-right)
• Look for ANY dark spot that doesn't match the surrounding sky or leaves
• Spend 20–40 seconds per tile minimum

Click EVERY spot you think contains a bird. Clicking roughly near the bird is fine.
• "Undo last click on this tile" — removes the most recent dot you placed on the current tile
• "← Previous tile" — go back to the tile you just left (you can add clicks you missed)
• "No birds visible" — only press this if you genuinely cannot see anything bird-like

Every tile in this set has at least one bird, so if you keep pressing "no birds visible" on every tile, the data will be useless. Take your time. Plan for ~60–90 minutes for ~96 tiles. Progress auto-saves; you can close and resume later.

🐦 Calibration — what a bird actually looks like at this scale

Before you start, look at these two reference tiles. The green circle marks a real bird (taken from a different photo in this dataset). The zoom-in shows the same bird at much higher magnification so you can see its shape.

Example 1 — bird ~39 px wide
cal1
Zoom-in (NN 8×)
zoom1
Example 2 — bird ~66 px wide (easy)
cal2
Zoom-in (NN 8×)
zoom2

What to look for in the annotation tiles: