6aa9c9bc07
Labeling: the six quality metrics in the plot menu, after Camera movement, as in Explore
maximeandClaude Opus 5.52026-11-04 12:33:36 +01:00
295c856ec2
Coherence of BREITHORN, MICHIGAN and NEBRASKA October labels (same building, per the expert)
maximeandClaude Opus 5.52026-11-04 09:28:19 +01:00
208db519ab
Labeler: lessons from the three October reviews (BREITHORN, MICHIGAN, NEBRASKA)
maximeandClaude Opus 5.52026-11-04 09:06:44 +01:00
c76c747989
MICHIGAN and NEBRASKA October labels reviewed by the analyst (/review-auto-labels), all suggestions applied
maximeandClaude Opus 5.52026-11-04 09:04:18 +01:00
cef7754888
Reviewed labels: "revue" is the review date, as for manual labels, and is not reviewed again
maximeandClaude Opus 5.52026-11-04 08:46:06 +01:00
83479499a9
Review of the labeler's labels: /review-auto-labels (D-128); BREITHORN October reviewed
maximeandClaude Opus 5.52026-11-04 08:40:27 +01:00
36c8cebccd
View: green arrows with a wide head; a placed range can be edited (drag an edge or the band)
maximeandClaude Opus 5.52026-11-03 18:26:56 +01:00
53aa95fd90
View: a range bar under the slider draws each sensor's displacement arrow (corr_x, corr_y end minus start)
maximeandClaude Opus 5.52026-11-03 18:23:18 +01:00
46125ff173
NEBRASKA presentation: clearer wording on the camera movement left in the curves
maximeandClaude Opus 5.52026-11-03 17:58:29 +01:00
31a846656c
BREITHORN presentation: vertical stability stated as a 0.2 px band
maximeandClaude Opus 5.52026-11-03 17:56:33 +01:00
fab7ddd9fd
Presentations: the lost reference is the message (MICHIGAN, NEBRASKA); BREITHORN parallax in corr_y
maximeandClaude Opus 5.52026-11-03 17:49:34 +01:00
85e590762b
Presentations for the client: Sandouville, no data-gap slide, settlements in corr_y (D-127)
maximeandClaude Opus 5.52026-11-03 17:36:25 +01:00
efbc5f15d7
Reading tools: what the etiqueteur had to script itself (metrics, daily bounds, local level, zone, references)
maximeandClaude Opus 5.52026-11-03 16:56:59 +01:00
6fbd7510b7
Etiqueteur: October labels for 8 stations; its folder moves to labeling/auto/etiqueteur/, analyses and scripts versioned (D-126)
maximeandClaude Opus 5.52026-11-03 16:33:41 +01:00
2bba388290
Detectors receive the station temperature; ISO dates accepted (D-124)
maximeandClaude Opus 5.52026-11-03 15:56:30 +01:00
daa1fd989d
Detector-learner: label bounds excluded, false positives as a regression, control watches the session station (D-123)
maximeandClaude Opus 5.52026-11-03 15:45:13 +01:00
e39531fd33
Detector-learner: user decisions apart in the bilan, process agents without events-learner (D-122)
maximeandClaude Opus 5.52026-11-03 15:21:29 +01:00
6d78bac736
Detector-learner: one commit per suggestion, retirer verdict, measured regression (D-121)
maximeandClaude Opus 5.52026-11-03 15:19:23 +01:00
e8e6cdd96d
JUNGFRAU labels get their family; lone jumps scored as Displacement (D-120)
maximeandClaude Opus 5.52026-11-03 15:16:21 +01:00
98c3308ca7
Detector-learner: one benchmark, scripts moved, critic with realistic suggestions (D-120)
maximeandClaude Opus 5.52026-11-03 15:12:24 +01:00
98be20a2bc
One detector per family, family keys equal to their names (D-119)
maximeandClaude Opus 5.52026-11-03 14:57:52 +01:00
0666d59e7d
Detector agents aligned on all labels of labeling/auto/, without events (D-118)
maximeandClaude Opus 5.52026-11-03 14:43:22 +01:00
516b538cf7
Labels: the trou family and corr_x/corr_y as the anomaly (D-111); thermal movement is always labeled, the camera's included even when compensated (D-117)
maximeandClaude Opus 5.52026-11-03 14:13:15 +01:00
0b1c93d135
Status page shows the labels of labeling/ (auto, else manual) instead of the detectors' events; compare.py: labels vs detectors by family, with the score, under the data loader (port 8769)
maximeandClaude Opus 5.52026-11-03 14:12:58 +01:00
97253639fa
Detector-learner: the global benchmark on all labeled stations at each step, written to detector-learner/historique.csv; the control reads it
maximeandClaude Opus 5.52026-11-03 14:01:55 +01:00
dbff2b947a
Detector-learner: regression.py is the global benchmark on all the labeler's labels, with its history (detector-learner/historique.csv) to compare sessions
maximeandClaude Opus 5.52026-11-03 14:01:32 +01:00
2a8d3da49d
Detector-learner: --ameliorer, the improver after the releveur as in /events-learner (D-071); the improver covers both processes
maximeandClaude Opus 5.52026-11-03 13:47:44 +01:00
0537f4ddeb
Detector-learner (D-116): the detectors learn from the labeler's labels alone, without events; critic, upgrader and control, the control guarded by the score on the other labeled stations (regression.py); NEBRASKA and NORDEND labels
maximeandClaude Opus 5.52026-11-03 13:46:36 +01:00
9724ea5c3b
Presentations: MICHIGAN and NEBRASKA say the redundancy is lost with two references and the correction no longer compensates "correctement"; MICHIGAN's two targets too close since the data start, jump on 5 Oct; sources compile in place in VS Code (brand/ link, Tinymist export on save)
maximeandClaude Opus 5.52026-11-03 12:29:37 +01:00
3211c9ce30
Presentations: their sources live next to each report (support/<dossier>/presentation.typ) and compile with .claude/skills/presentation/compiler.sh
maximeandClaude Opus 5.52026-11-03 12:03:03 +01:00
1a04ae0936
Presentations: BREITHORN slide 8 shows the camera's rotation above the targets' curves (the rotation drops on each of the three nights, then comes back); conclusion slides titled "Conclusions" with the bullet points
maximeandClaude Opus 5.52026-11-03 11:58:56 +01:00
44711d2288
Presentations: camera tilt and pan, captions that read on their own, conclusions as bullet points without any instruction to the client
maximeandClaude Opus 5.52026-11-03 11:57:31 +01:00
a85d736714
Analyst reads the data like the labeler (D-115): same tools and documentation (labels map, families, measurement chain, temperature from metrics/), in analyser-lot and the /labels-learner instructions; its output (etiquettes.json, journal categories, bounds) unchanged for the detectors benchmark; analyser-lot no longer calls itself the loop's truth (D-112)
maximeandClaude Opus 5.52026-11-03 11:44:00 +01:00
05e7494e64
Presentations: facts only, nothing asserted about the structure (user: we do not know whether it moves); "Point de situation" on the title page; the camera's movement is real, its cause (fixing, wall or other) not known
maximeandClaude Opus 5.52026-11-03 11:40:46 +01:00
98261833f6
Support: situation reports for BREITHORN, MICHIGAN and NEBRASKA extended to the whole of October 2026, and their presentations; new /presentation skill (D-114)
maximeandClaude Opus 5.52026-11-03 11:31:17 +01:00
ae543e921d
Support: MICHIGAN and NEBRASKA situation reports checked against the completed labels, with the expert's answers. MICHIGAN: COLH21' and REFH22' are so close their spots partly overlap, move one; the curves drop when the station's temperature falls and the camera tilts, corrected on two references only (factory, no day/night cycle). NEBRASKA: the camera slowly tilts nose down, the client checks its fixing; REFCOLH22 lost at 12:00 on 09.10; three mornings where the curves jump; temperature not a cause (indoors)
maximeandClaude Opus 5.52026-11-03 09:51:03 +01:00
f0ba5b9fcf
Labels and journal: "Faulty target" renamed "Target issue" everywhere (user: the name was not clear), key cible_degradee -> cible_a_probleme; labels, events, detectors, Status page, labeling tool, documentation and history (D-109)
maximeandClaude Opus 5.52026-11-03 09:31:18 +01:00
6955adf6f5
Labeler on MICHIGAN and NEBRASKA, October 2026, with the expert's diagnosis: one reference lost on 09.10 (big jump, compensation on two references afterwards), slow camera drift with the falling temperature, brief thermal deviations (MICHIGAN: indoor factory, no day/night effect); replaces the previous labels from 20.10
maximeandClaude Opus 5.52026-11-03 09:24:52 +01:00
350663bd13
Support: BREITHORN situation report checked against the completed labels: COLI24 drifts the other way in corr_y with the camera tilt, phases of the drift, nights where the camera moves a little more (not thermal), one reference hidden at a time on the nights of 25 to 27.10, by a vehicle that looks like a pallet carrier (expert)
maximeandClaude Opus 5.52026-11-03 09:09:46 +01:00
b4166605b5
Labeler on BREITHORN, October 2026: 26 labels in labeling/auto/ (17 obstructions by a vehicle, the camera's drift in four episodes with the expert's diagnosis: the camera tilts down and its movement is not fully compensated, the targets do not move; four brief camera deviations; one gap); camera tilt sign documented (positive = nose down); labeling tool opens both plots on corr_x and corr_y, labels colored by family
maximeandClaude Opus 5.52026-11-03 08:34:03 +01:00
262af9fdf4
Support: BREITHORN situation report rewritten with the expert's diagnosis: the camera slowly tilts nose down and the compensation corrects some targets badly (parallax), so POUI19 and COLJ19 drift while the points do not move; a vehicle hides the low targets on several nights; the night variations are not thermal (stable building), cause not established
maximeandClaude Opus 5.52026-11-03 08:31:54 +01:00
fddf0cda88
Labeling tool: two stacked plots, each with its own menu, sharing time range and margins with the label bar below; the Labels menu opens on auto
maximeandClaude Sonnet 5.52026-11-03 08:31:13 +01:00
026a5bd432
Support skill: /support <STATION> writes a situation report from the labeler's labels (labeling/auto/), one or two images per finding (curves of the targets, camera image when the cause shows), one section per slide; /support <mail> unchanged (D-113)
maximeandClaude Opus 5.52026-11-02 17:45:36 +01:00
30940e4f50
Support: situation reports for BREITHORN, MICHIGAN and NEBRASKA (20.10 to 01.11.2026), from the labeler's labels, with curves and camera images (D-113)
maximeandClaude Opus 5.52026-11-02 17:44:39 +01:00
bc984a86d3
Events-learner on the labeler's labels: the session reads the lot's labels in labeling/auto/ instead of having the analyst read the lot; the detectors-critic judges the detectors against them, family by family (correspondence of labeling/familles.md); score.py --du --au scores by family and by instant, the session page shows labels and detectors on one row per family; analyst, tool-builder and event-builder leave the session (D-112, point 72)
maximeandClaude Opus 5.52026-11-02 16:25:44 +01:00
a95d7ca07a
Labeling tool: a one-measurement label keeps a visible width on the bar and the plots (0.6 % of the page period, at least 15 min); at 15 min the white border hid it
maximeandClaude Opus 5.52026-11-02 15:46:35 +01:00
5adb711ce0
Label families: five families for labels, documented in labeling/familles.md and read by the labeler (Visibility, Obstruction, Thermal displacement, Displacement, Faulty target; turbulence merged into Visibility) (D-109, D-110); the labeler keeps only its labels, its work folder goes to out/ (D-106); third trial on WILDSTRUBEL 01-06.09, 30 labels; the labeling tool shows one-measurement labels
maximeandClaude Opus 5.52026-11-02 15:40:35 +01:00
8266293ed3
Labeler agent: an analyst that only produces labels (etiqueteur, D-106), one category per label (D-107), anomalies only and one Thermal displacement category (D-108); second trial on JUNGFRAU 30.09-05.10, 15 labels; the labeling tool gets a Category box above the text, to add and to edit
maximeandClaude Opus 5.52026-11-02 14:44:08 +01:00
fe025612cc
Automatic labels, first trial: the analyst labels JUNGFRAU 20-25.10 alone with the labels map (7 labels in labeling/auto/, analysis beside them); the labeling tool shows manual or auto labels from a Labels menu (D-105)
maximeandClaude Opus 5.52026-11-02 13:40:29 +01:00
d47d29b04e
Labels map: the reviewed expert labels arranged by phenomenon for the analyst (labeling/carte-des-etiquettes.md), built by the /labels-map skill with its check script; first pass on 23 labels, 14 phenomena, reviewed against the labels and the skill strengthened (D-104)
maximeandClaude Opus 5.52026-11-02 12:47:15 +01:00
d3b89a7f2e
WILDSTRUBEL and FLORIDA labels reviewed with the expert (/review-labels): the expert's own label and the twin targets kept, duplicates and a contradicted one removed; 24 labels in all, reviewed except PS_DES_COUDRES (no data)
maximeandClaude Opus 5.52026-11-02 12:28:08 +01:00
a36648c836
JUNGFRAU labels reviewed with the expert (/review-labels): all 8 kept with their bounds, rewritten for clarity; the skill keeps the expert's own labels and only asks on a real objection
maximeandClaude Opus 5.52026-11-02 12:24:34 +01:00
6846976c2e
CLARIDEN labels reviewed with the expert (/review-labels): 8 merged into 2 (the occluded target, the real movement of the structure), plain qualitative words
maximeandClaude Opus 5.52026-11-02 12:15:30 +01:00
46a8d5a22c
NORDEND labels reviewed with the expert (/review-labels): 6 kept, rewritten in plain qualitative words; 5 minor ones removed
maximeandClaude Opus 5.52026-11-02 11:57:36 +01:00
7ca701e502
Label review as the /review-labels skill, one label at a time with the expert, a qualitative text proposed for each, no review file (D-103); NIEDERHORN reviewed: 5 labels kept, 3 train passages removed
maximeandClaude Opus 5.52026-11-02 11:26:34 +01:00
c72ef6fdf1
Labeling tool: the open label's bounds are edited in two UTC date fields and saved with its text; plan point 71, an integrator for the labels
maximeandClaude Opus 5.52026-11-02 10:40:24 +01:00
3a07d45ffe
Labels in two folders: manual/ (by hand, checked with the label-reviewer, the tool's) and auto/ (what the analyst will find alone, empty) (D-102)
maximeandClaude Opus 5.52026-11-02 10:27:32 +01:00
148d197173
Labeling tool: a white outline separates adjacent labels on the bar
maximeandClaude Opus 5.52026-11-02 10:01:46 +01:00
96f4bb58aa
Labeling tool: Delete removes a label (red button, right of Save and Close), hover shows sensor, value and UTC time, text boxes 10 lines high
maximeandClaude Opus 5.52026-11-02 09:56:04 +01:00
3ce582d8b5
run.sh: the three workstation tools (Status, View, Labeling) in one command, each by its data loader
maximeandClaude Opus 5.52026-11-02 09:47:07 +01:00
63c9f9e625
Labels: the validated events' diagnoses become labels in their stations' files, real sensor names, same format as the expert's (D-100)
maximeandClaude Opus 5.52026-11-02 09:42:08 +01:00
713bb80aa3
Label-reviewer: an agent checks the expert's labels against the data and asks before any change (D-101); labels as the knowledge base, validated events get theirs (D-100)
maximeandClaude Opus 5.52026-11-02 09:35:28 +01:00
d1fd7f6bdd
Labels-learner: the analyst measures the expert's labels and asks until it understands, a documentaliste writes its conclusions into the docs (D-099, point 69)
maximeandClaude Opus 5.52026-11-02 09:12:53 +01:00
010798d992
Labeling tool: a click on a label shows it, its text is edited and saved (point 68)
maximeandClaude Opus 5.52026-11-02 08:56:16 +01:00
b52f1c0d22
View tool: Labels checkbox, sensor names on the images, targets and references apart
maximeandClaude Opus 5.52026-10-30 14:18:14 +01:00
83f7f7be7f
Labeling tool: system metrics in the plot menu, identifiers left out; point 69: how the analyst learns from labels, still open
maximeandClaude Opus 5.52026-10-30 11:23:31 +01:00
fcd9191a95
Labeling tool: the expert labels ranges in plain text, one plot at a time (D-098, point 68)
maximeandClaude Opus 5.52026-10-30 08:59:38 +01:00
7fb176cf0a
Events-learner: the analyst writes its own tools, a tool-builder brings them into agent-tools; /aiguiser retired (D-097)
maximeandClaude Opus 5.52026-10-29 17:10:17 +01:00