lowestcase-glyphs-sdf

Glyph dataset for recreating Tom 7's "Lowestcase and uppestcase letters: Advances in derp learning" (SIGBOVIK 2021, http://tom7.org/lowercase/).

Each of 320 Google Fonts (regular weight) contributes 26 uppercase and 26 lowercase glyph images, rendered as 64×64 signed distance fields (each pixel = signed distance to the glyph edge, clamped at ±8 px and mapped to [0, 1], so 0.5 = on-edge, > 0.5 = inside ink). Glyphs are normalized to a 48×48 active area, centered.

Contents

sdf_glyphs.npz:

• up: float16 (320, 26, 64, 64) — uppercase A–Z per font
• lo: float16 (320, 26, 64, 64) — lowercase a–z per font
• fonts: font family names
• ious: per-font mean IoU between normalized uppercase and lowercase glyphs

preview_*.png: eyeball strips (rows: uppercase / lowercase; columns A G L S Z).

Filtering

Fonts are rejected if any of the 52 glyphs is missing/blank, or if uppercase ≈ lowercase after normalization (mean IoU > 0.90 — small-caps and uppercase-only fonts make the mapping degenerate). 320 of ~354 tried fonts survived; kept-font IoUs are far below the threshold.

Source

Original experiment: Tom Murphy VII, Lowestcase and uppestcase letters: Advances in derp learning, SIGBOVIK 2021 — http://tom7.org/lowercase/lowercase.pdf (video: https://youtu.be/HLRdruqQfRk). Fonts via the Google Fonts catalogue (gwfh.mranftl.com API); check the OFL/Apache license of individual font families before redistribution of rendered outlines.