test: add real-photo fixture harness for detection regressions

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
immortaly007andClaude Opus 5.5 committed 2026-10-15 14:45:10 +02:00
1 parent fad7deed5d
commit 0a97cb88af
3 files changed
+73

No files matched your search

+27
View File
@@ -0,0 +1,27 @@
# Real photo fixtures
Regression photos for `detect()`. Each entry in `manifest.json` is checked by `tests/fixtures.test.ts`.
To add one:
1. Take a top-down photo of a scrap piece with something of known length visible.
2. Rotate it upright and resize it so the long side is at most 2000 px. Save as JPEG.
EXIF orientation is ignored by the test decoder.
3. Measure the real piece's bounding box in mm.
4. Open the photo in the app, set the scale, and note px/mm, the threshold you used, and,
if needed, a seed point (shown in image px).
5. Add an entry:
```json
{
"file": "plywood-1.jpg",
"threshold": null,
"invert": false,
"seed": null,
"pxPerMm": 3.21,
"expected": { "widthMm": 412, "heightMm": 287, "holes": 2 },
"toleranceMm": 8
}
```
`threshold: null` means Otsu's automatic value. `seed` is `{ "x": ..., "y": ... }` or `null`.
+1
View File
@@ -0,0 +1 @@
[]