auto: the long labels requalified as analyses like the others, <STATION>_<start>_<end>/ with request.json (ex long-labels/<STATION>/); their labels carry the analysis; work/ out of git
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
1 parent
ddb17c2189
commit
764af7d8ec
132 files changed
+125
-9256
No files matched your search
@@ -0,0 +1,44 @@
|
||||
"""Write labels.jsonl of the NORDEND long-label requalification: 24 and 152 kept as they are, 169 and 195 corrected."""
|
||||
import json
|
||||
import os
|
||||
|
||||
here = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
|
||||
labels = {d["id"]: d for d in map(json.loads, open(os.path.join(here, "input.jsonl")))}
|
||||
|
||||
d169 = labels["NORDEND-169"]
|
||||
d169.pop("review")
|
||||
d169["end"] = "2026-11-07T14:00Z"
|
||||
d169["description"] = (
|
||||
"PR17: a blurred pale disc (mark on the window lit by the infrared) sharpens just under its spot at 02:30 and "
|
||||
"covers part of it (thumbnails of 08/10 at 02:00 and 02:45, then of the following nights; still visible, weaker, "
|
||||
"on 26, 29 and 31/10, and as a pale area under the spot on 04/11). In one step, its corr_y drops and its corr_x "
|
||||
"rises markedly, only PR17 moves, its neighbours do not, and its error_value doubles and stays high. This step "
|
||||
"brings the corr_y of PR17 back to roughly its level from before its jump of 05/09 (separate label, Displacement, "
|
||||
"plateau until that day); but the September jump had only touched its corr_y, without raising its error_value, "
|
||||
"and no disc was visible then: the step comes from the mark, and that it also undoes the September plateau is not "
|
||||
"established. Its curves still change level in steps: a partial return at 23:00 on the 12th, in the middle of a "
|
||||
"back-and-forth of the camera from 22:30 to 00:15 (labelled separately, Displacement) that moves the other "
|
||||
"targets too; they come back after it, PR17 keeps its new level, so its step is its own and not the camera; then "
|
||||
"new offsets from 21 to 24. After the rain of 24 to 25, its error_value falls back to its earlier level, but its "
|
||||
"curves come back neither to their level of the beginning of the month (the September plateau) nor to that from "
|
||||
"before the jump of 05/09, and stay so until the end of the range, the disc still visible. Cause seen on the "
|
||||
"images; the exact nature of the mark is not established. This is not a movement of PR17; ignore its curve over "
|
||||
"the range, clean the window.")
|
||||
|
||||
d195 = labels["NORDEND-195"]
|
||||
d195.pop("review")
|
||||
d195["end"] = "2026-11-07T14:00Z"
|
||||
d195["description"] = (
|
||||
"CTR01 (local reference, read as a target): from the evening of the 16th, its error_value no longer falls back "
|
||||
"to its level of the beginning of the month and keeps rising until the end of the month, very close to the "
|
||||
"threshold then above it; the measurement is rejected a large part of the time, by night as by day, every day, "
|
||||
"almost without stopping from the end of October to the afternoon of 02/11, then still about half of the time "
|
||||
"until the end of the range; its curve is missing or stays very noisy. Before the 16th, it was only rejected on "
|
||||
"sunny afternoons and during the nights when the left-edge spots fade out (separate labels). On the images, its "
|
||||
"spot is weak; at the end of the month, blurred marks left on the window after the rain pass near it (26/10 image "
|
||||
"at 03:00), then a pale blurred band covers it (31/10 image at 03:00). Cause not established: state of the prism, "
|
||||
"or of the window in front of it, to be checked. Ignore its curve; intervene on the target.")
|
||||
|
||||
with open(os.path.join(here, "labels.jsonl"), "w") as f:
|
||||
for k in ("NORDEND-24", "NORDEND-152", "NORDEND-169", "NORDEND-195"):
|
||||
f.write(json.dumps(labels[k], ensure_ascii=False) + "\n")
|
||||
Reference in new issue
Block a user