dandri/ESLPwn
1
0
Fork 0
mirror of https://github.com/i12bp8/ESLPwn.git synced 2026-09-25 12:53:36 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity
44 Commits 2 Branches 2 Tags
3ebae10adaa1b9daef919dddfc5bcae5efcb1062
Commit Graph
4 Commits
Author SHA1 Message Date
sosnekandCursor b328511d02 Add SmartTAG Color 2.6 (type 1626) support 
These tags look like 296x152 but the firmware only takes 152x296, so landscape BMPs get transposed. Tested on a real 2.6 color tag.

Co-authored-by: Cursor <cursoragent@cursor.com>
 2026-08-31 00:07:03 -04:00
i12bp8 99bcc162b8 web-image-prep: chroma-contracted quantiser for newsprint illusion
Deploy Image Prep to GitHub Pages / deploy (push) Failing after 31s
Details 
Previous version pulled accent ink too aggressively (accentPull 0.85)
so anything red-leaning collapsed to solid red. Real photographic feel
from a 3-ink palette comes from W/B halftones simulating greys with
the chromatic ink only sparsely overlaid - the newsprint illusion.

Three knobs now make this work:
  - L_WEIGHT 3.5 (was 1.6): lightness completely dominates the metric
  - CHROMA_CAP: source a/b axes pre-scaled to the achievable accent
    chroma so the dither isn't asked for more colour than one ink can
    deliver. Excess chroma gets dropped via FS error diffusion through
    B+W neighbours.
  - accentPull 0.06 (was 0.85): only a tiebreaker bonus, used in the
    most strongly hue-aligned pixels.
 2026-04-26 21:03:10 +02:00
i12bp8 ad9b6dd2fe web-image-prep: hue-aligned tri-colour quantiser + serpentine dither 
The old quantiser pulled accent ink toward any chromatic pixel using a
flat chroma bonus, which smeared red across shadows and produced flat
red blobs without depth. Two fixes:

- Distance now weights L (lightness) ~1.6x more than chroma, so dark
  pixels go to black and light pixels to white before colour competes.
  That preserves the photograph's tonal structure.
- Accent attraction is the squared *signed projection* of the pixel's
  chroma onto the accent direction. Red ink only pulls red-leaning
  pixels (positive a*); blue/green never picks red. The squaring gives
  a smooth ramp, so faint reds get sparse red dither and saturated
  reds get dense red - matching how a real risograph layers a single
  chromatic ink under W/B halftones.

Also flipped Floyd-Steinberg to a serpentine scan with mirrored kernel
on reverse rows, killing the diagonal worm artefacts that gave smooth
gradients an obviously-dithered look.
 2026-04-26 20:59:15 +02:00
i12bp8 d608afb43b Drop-folder image flow + web image preparer 
- Add web-image-prep/: self-contained HTML5 image preparer with
  perceptual Oklab 3-colour quantisation, Floyd-Steinberg/Atkinson/
  Bayer dither, all supported tag profiles, BMPs that match the FAP
  reader exactly. Auto-deployed to GitHub Pages.
- Add 'Set Image' under Targeted Payloads. Lists every BMP in
  apps_data/tagtinker/dropped/, scrolls long names, marks foreign-
  resolution files with a '~' prefix.
- Stream-time nearest-neighbour rescaling so any BMP can target any
  tag at any page; image options screen is now Page-only.
- Remove Phone Sync menu and the Android companion app: replaced by
  the browser preparer + drop folder.
- Static, IR-friendly transmit screen ('Tinkering Tag', point hint,
  3-dot loader, 'Flipped ;)').
 2026-04-26 15:40:27 +02:00