add per-cu3s README: 2026_03_03_11-31-31_README.md
Browse files
data/day2/2026_03_03_11-31-31_README.md
ADDED
|
@@ -0,0 +1,64 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# 2026_03_03_11-31-31
|
| 2 |
+
|
| 3 |
+
**Capture date:** 2026-03-03 · **Day:** `day2` · **Subfolder timestamp:** `2026_03_03_11-31-31`
|
| 4 |
+
|
| 5 |
+
**Operator:** [Anish Raj](mailto:raj@cubert-gmbh.de) — Cubert GmbH
|
| 6 |
+
|
| 7 |
+
## Hardware
|
| 8 |
+
|
| 9 |
+
- **Camera:** Cubert Ultris XMR (serial 254902)
|
| 10 |
+
- **Lens:** 50 mm / f2.0
|
| 11 |
+
- **Accessories:** None
|
| 12 |
+
|
| 13 |
+
## Setup & Lighting
|
| 14 |
+
|
| 15 |
+
- **Light source:** Halogen (4 lights in 4 configurations)
|
| 16 |
+
- **Working distance:** 46.6 cm (466 mm)
|
| 17 |
+
- **Field of view:** ≈12.5 × 12 cm
|
| 18 |
+
- **Notes:** Four lighting configurations used per arrangement:
|
| 19 |
+
- `l0` — all four lights on
|
| 20 |
+
- `l1` — front and back lights only
|
| 21 |
+
- `l2` — front and two side lights
|
| 22 |
+
- `l3` — front light only
|
| 23 |
+
|
| 24 |
+
All four frames of the same scene arrangement form a `group_id` in `splits.csv`
|
| 25 |
+
and are always kept in the same train / val / test split.
|
| 26 |
+
|
| 27 |
+
## Acquisition Settings
|
| 28 |
+
|
| 29 |
+
- **Measurement mode:** Snapshot (single capture per lighting config)
|
| 30 |
+
- **Integration time:** 15 ms
|
| 31 |
+
- **White reference:** 55 % gray target
|
| 32 |
+
- **Dark current:** Recorded
|
| 33 |
+
- **Processing mode at rest:** `Preview` — call `cuvis.ProcessingContext(sf).processing_mode = cuvis.ProcessingMode.Reflectance; ctx.apply(m)` before reading `m.cube`
|
| 34 |
+
|
| 35 |
+
## Frames in this `.cu3s`
|
| 36 |
+
|
| 37 |
+
| | |
|
| 38 |
+
|---|---:|
|
| 39 |
+
| Total saved frames | **17** |
|
| 40 |
+
| Frames with foreign-object annotations | **0** |
|
| 41 |
+
| Train / val / test split | **10 / 3 / 4** |
|
| 42 |
+
| `global_frame_id` range (day) | `2 .. 18` |
|
| 43 |
+
| `local_image_id` range (this cu3s) | `0 .. 16` |
|
| 44 |
+
|
| 45 |
+
## Annotations
|
| 46 |
+
|
| 47 |
+
| Class | Frame-level count |
|
| 48 |
+
|---|---:|
|
| 49 |
+
|
| 50 |
+
|
| 51 |
+
Annotations are semantic-mask COCO polygons. See [`2026_03_03_11-31-31.json`](./2026_03_03_11-31-31.json).
|
| 52 |
+
|
| 53 |
+
## Scene description
|
| 54 |
+
|
| 55 |
+
Mixed beluga + dark green marbled lentils (Emershofer brand) arranged on the
|
| 56 |
+
conveyor-belt material. Foreign objects placed by hand into the bulk product.
|
| 57 |
+
For each arrangement, four snapshots are captured (one per lighting config) to
|
| 58 |
+
study how illumination affects spectral separability.
|
| 59 |
+
|
| 60 |
+
## Credits
|
| 61 |
+
|
| 62 |
+
Recorded and processed by the AI Team @ Cubert. See the top-level
|
| 63 |
+
[`README.md`](../../README.md) for the full dataset description, load snippets,
|
| 64 |
+
example frames, and split methodology.
|