Files
aimsight-labeling/auto/analyses/WILDSTRUBEL_2026-10-06T1945_2026-10-20T0000/analysis.md
T

46 lines
5.3 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
# Labels: WILDSTRUBEL, 2026-10-06T19:45Z → 2026-10-20T00:00Z (UTC)
**Incomplete verification**: this session hit the usage limit. I checked the curves, the rigid analysis and the metrics. I did **not** look at the images (thumbnails) or run the precedents. 22 labels are kept as they were, 1 is corrected, none are removed or added. The open points are listed in section 2.
## Verdict per old id
| id | Verdict | Evidence |
|---|---|---|
| WILDSTRUBEL-176 | kept | The gap runs from 06/10 19:45 to 07/10 07:00, with isolated measurements at 05:00, 06:00 and 06:30 (inventory). Before the gap the voltage is low and the battery at about 22 %. The metrics stop from 21h to 03h (metrics.py). |
| WILDSTRUBEL-177 | kept | Tilt rises from 05:00 (0.149) to 06:30 (0.185), falls to 0.108 at 11:15, and is back near 0.134 at 12:00. Rotation is back at 12:30. The description already says so, and the gap with the bound is within the uncertainty. |
| WILDSTRUBEL-178 | kept | Daytime agitation of the corrected curves and sunrise rejections, as the description says. Open point: PA_9 is rejected at 10:45 and 11:00 and no label names it (section 2). |
| WILDSTRUBEL-179 | kept | At 11:30, PA3_2, PB4, PB_5, PA1_2 and R2 are rejected. PA_9 and PB2_2 move about three noises, which the text calls "barely". |
| WILDSTRUBEL-180 | kept | Rejections at 11:45 (R1), 12:00 (R1, R2, PB4) and 12:15 (PA3_2). Note: at 11:45, PB4, PB_5 and PB3_2 already jump with R1 out. 11:45 is touched and is the shared bound with 179. Thumbnails not looked at. |
| WILDSTRUBEL-181 | kept | At 12:45, R2 and PB4 are rejected. |
| WILDSTRUBEL-182 | kept | Measurements stop from 13:30, except one at 14:15. The battery drops to 22 % and the voltage to 12.40 at 14h. |
| WILDSTRUBEL-183 | kept | corr_x of PB1_2 goes from about -0.04 (07/10 afternoon) to -0.20/-0.26. PB2_2 goes from -0.23 to -0.33/-0.38. With base 07/10 13:00, their rigid residual dx is about -0.1 px against the block (the other targets are near 0). |
| WILDSTRUBEL-184 | kept | Rotation drops from -0.05 to -0.11 between 09:15 and 13:00 and holds on 09/10 with six references. The rigid fit of the targets also turns by about -0.04°, so the camera itself rotates. |
| WILDSTRUBEL-185 | kept | Sunrise rejections from 05:30 to 07:15. The daytime agitation is there. Open point: isolated rejections at 08:30 (PA_9), 08:45 (PA3_2) and 09:00 (PB2_2) are not covered by any separate label. |
| WILDSTRUBEL-186 | kept | PA1_2 is rejected from 06:15 to 11:00, and its returns at 11:15 and 12:30 bring the targets back up (corr_x table). |
| WILDSTRUBEL-187, 188, 189 | kept | The rejections match the inventory. Thumbnails not looked at again. |
| WILDSTRUBEL-190, 191 | kept | At 14:15 and 14:45, R1 and R2 are rejected and PB_5 and PB1_2 jump. |
| WILDSTRUBEL-192 | kept | The gap runs from 16:00 to 05:00 with the battery at 24 % and the voltage at 12.50. |
| WILDSTRUBEL-193, 194 | kept | On 09/10 from 05:00 to 05:30, no reference is kept (inventory). The tilt rises and comes back. |
| WILDSTRUBEL-195 | kept | At 09:30, PB_5 and PB3_2 are rejected. |
| WILDSTRUBEL-196 | kept | At 12:00, corr of PA3_2 and PB4 is far off (both axes), and R2 is rejected. |
| WILDSTRUBEL-197 | kept | On 09/10 at noon the battery is at 67 % and the voltage at 13.03, then nothing until 13/10 03h. After that the metrics are back, the battery is full by day and stays near 80 % at night, and there are hourly images but no measurements. |
| WILDSTRUBEL-198 | **corrected** | The raw rows contradict "all the references are rejected": R1 is accepted (error_value below its threshold, corr = 0), and the other five are rejected. All the well-found sensors have relative_y about 12 px higher than on 09/10 12:30, and x about 2 px: the camera moved during the outage. This is not in the old text. The family is kept: I did not re-check the image. |
Spanning labels: WILDSTRUBEL-130 (R3 rejected by day) holds within the range. WILDSTRUBEL-199 holds from 15/10 08:00: no metrics from 15/10 09h to 20/10. Note: by the rule "two gaps separated by a single measurement make one No data", 197 and 199 should be one label. I leave the split because the cause differs (metrics present on 13–15/10).
## 1. My labels
The 23 labels of `labels.jsonl` are the existing ones; only 198 changed (see above).
## 2. What I left without a label (open, not verified)
- 07/10 10:45–11:00: PA_9 rejected twice. 08/10 08:30, 08:45 and 09:00: isolated rejections (PA_9, PA3_2, PB2_2). At 08:30 the pose jumps (tilt 0.224) and PB1_2 dips. 09/10: the references PA1_2 and PA_8 drop out at times (08:00–08:30, 09:15, 09:45, 10:15–10:30, 11:30), and PA2_2 and PB_5 move with them. These may each need their own label (passage or reference loss). The thumbnails need checking.
- The base change between 07/10 and 08/10 (R1 dy about -0.3 px against the block of targets) moves PA2_2 and PB_5 in corr_x by about two noises. I did not label it.
## 3. The documentation
Rules and families were clear. No weather folder exists for WILDSTRUBEL.
## Difficulties
- Usage limit reached before the image review and the precedents.
## Sources read
tools/GUIDE.md, tools/DATA.md, labeling/families.md, labeling-rules.md, reading-rules.md, stations.md, labeling/auto/WILDSTRUBEL.jsonl (labels around the range), work/existing.jsonl, work/spanning.jsonl; inventory, overview (with --rigid), summary and metrics outputs in work/; raw rows of 09/10 12:30 and 15/10 08:00.