Files
aimsight-labeling/auto/analyses/NEBRASKA_2026-10-09T0800_2026-11-01T0000/report.md
T

8 lines
641 B
Markdown

# NEBRASKA: requalified long labels
2 labels: 1 kept, 1 corrected, 0 removed.
- **NEBRASKA-6** (corrected): the drift of 16 October → 1 November is clearly above the threshold (up to about 17 noises on COLJ20; in mm, about +1.2 mm on COLJ20, +1.0 on POUJ19, +1.5 on COLJ24). Only the description changed: the tool name `ensemble.py --rigide`, which does not exist, is replaced by what it did, and the description now says that COLJ24, far from the camera, drifts as much as the others in mm.
NEBRASKA-2 is kept: REFCOLH22 is rejected at 100 % of its measurements after 9 October, and its loss makes the targets jump by up to 13 noises.