Auto labels reviewed per station and cross-reviewed; all labeling knowledge in labeling/ (D-130)
- Every station's auto labels reviewed by an analyst (/review-auto-labels), then a cross-review per family across all stations (new skill /cross-review-labels), a finishing pass, and descriptions rewritten in words (no numeric values). - The three never-labeled ranges (WILDSTRUBEL 07/09-01/10, NORDEND 16/08-01/10) labeled and reviewed: 991 auto labels on 11 stations, labeling/controle.py at 0 errors. - Labeling knowledge gathered in labeling/: regles-d-etiquetage.md (new), familles.md (settled boundaries), connaissances/ (notes moved from docs/, links updated), the label map rebuilt from reviewed labels, events and cases; events tied to their stations (PS_DES_COUDRES is JUNGFRAU). The labeler and reviewers read only labeling/. - labels-map reads auto labels and events; decision D-130; living plan points 73-75. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
1 parent
8e1cfc52c9
commit
35b5ed4d74
298 files changed
+30413
-960
No files matched your search
@@ -0,0 +1,71 @@
|
||||
"""Frontière Visibility / Thermal displacement : sur la plage d'une étiquette (et une marge), mesure par mesure,
|
||||
ce qui bouge avec les courbes : l'error_value des cibles (toute la scène ?), la température du boîtier,
|
||||
l'échelle et la rotation de la caméra.
|
||||
|
||||
Niveaux : médiane des 24 h qui finissent 2 h avant le début. Colonnes : heure ; n_corr = cibles dont corr_x ou
|
||||
corr_y sort de 6 MAD de leur niveau ; ev = médiane sur les cibles de error_value moins niveau ; n_ev = cibles dont
|
||||
l'error_value sort de 6 MAD ; temp = température du boîtier (heure la plus proche) moins niveau ; scale, rot = pose
|
||||
moins niveau, en MAD de la pose.
|
||||
|
||||
Usage (depuis la racine) : python3 frontiere.py ID [--marge 2] [--pas N] (une ligne toutes les N mesures)
|
||||
"""
|
||||
import sys, json, statistics as st
|
||||
from datetime import timedelta
|
||||
sys.path.insert(0, "agent-tools")
|
||||
import donnees as d
|
||||
|
||||
a = sys.argv[1:]
|
||||
i = a[0]
|
||||
marge = timedelta(hours=float(a[a.index("--marge") + 1]) if "--marge" in a else 2)
|
||||
pas = int(a[a.index("--pas") + 1]) if "--pas" in a else 1
|
||||
nom, n = i.split("#")
|
||||
e = json.loads(open(f"labeling/auto/{nom}.jsonl").readlines()[int(n) - 1])
|
||||
t1, t2 = d.instant(e["debut"]), d.instant(e["fin"])
|
||||
dos = f"data/{nom}"
|
||||
par = d.charger(dos)
|
||||
cib = {s: ms for s, ms in par.items() if not d.est_reference(ms)}
|
||||
r0, r1 = t1 - marge - timedelta(hours=24), t1 - marge
|
||||
|
||||
|
||||
def niv(vals):
|
||||
if len(vals) < 4:
|
||||
return None
|
||||
m = st.median(vals)
|
||||
return m, max(st.median(abs(v - m) for v in vals), 1e-6)
|
||||
|
||||
|
||||
nv = {}
|
||||
for s, ms in cib.items():
|
||||
w = d.filtrer(ms, r0, r1)
|
||||
nv[s] = {q: niv([m[q] for m in w if m[q] is not None]) for q in ("corr_x", "corr_y", "error_value")}
|
||||
cam = d.charger_camera(dos)
|
||||
ncam = {q: niv([p[q] for p in cam if r0 <= p["t"] <= r1 and p[q] is not None]) for q in ("scale", "rotation")}
|
||||
pcam = {p["t"].replace(second=0, microsecond=0): p for p in cam}
|
||||
tmp = d.charger_temperature(dos)
|
||||
ntmp = niv([v for t, v in tmp if r0 <= t <= r1])
|
||||
idx = {s: {m["t"]: m for m in ms} for s, ms in cib.items()}
|
||||
inst = sorted({t for s in idx for t in idx[s] if t1 - marge <= t <= t2 + marge})
|
||||
print(f"{i} {e['debut']} -> {e['fin']} {e['categorie']}")
|
||||
print("heure n_corr ev n_ev temp scale rot")
|
||||
for k, t in enumerate(inst):
|
||||
if k % pas:
|
||||
continue
|
||||
nc = ne = 0
|
||||
evs = []
|
||||
for s in idx:
|
||||
m = idx[s].get(t)
|
||||
if m is None:
|
||||
continue
|
||||
L = nv[s]
|
||||
if m["error_value"] is not None and L["error_value"]:
|
||||
evs.append(m["error_value"] - L["error_value"][0])
|
||||
ne += abs(m["error_value"] - L["error_value"][0]) > 6 * L["error_value"][1]
|
||||
if m["corr_x"] is not None and L["corr_x"] and L["corr_y"]:
|
||||
nc += any(abs(m[q] - L[q][0]) > 6 * L[q][1] for q in ("corr_x", "corr_y"))
|
||||
tt = min(tmp, key=lambda x: abs(x[0] - t)) if tmp else None
|
||||
dt = f"{tt[1] - ntmp[0]:+5.1f}" if tt and ntmp and abs(tt[0] - t) < timedelta(hours=1) else " . "
|
||||
p = pcam.get(t.replace(second=0, microsecond=0))
|
||||
cs = " ".join(f"{(p[q] - ncam[q][0]) / ncam[q][1]:+6.1f}" if p and p[q] is not None and ncam[q] else " ."
|
||||
for q in ("scale", "rotation"))
|
||||
mark = "*" if t1 <= t <= t2 else " "
|
||||
print(f"{t:%d %H:%M}{mark} {nc:3d} {st.median(evs) if evs else float('nan'):+.4f} {ne:3d} {dt} {cs}")
|
||||
Reference in new issue
Block a user