auto: the analysts verified the labels of 32 periods (11 stations, up to 14 days each): 710 labels -> 835, 587 unchanged; one analysis folder per period with its report (section Verification)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
1 parent
df979301b5
commit
2a5d87659b
279 files changed
+10758
-710
No files matched your search
@@ -0,0 +1,51 @@
|
||||
"""Build labels.jsonl: the existing labels of the range, kept as they are (without id) or rewritten, plus the added ones."""
|
||||
import json, os
|
||||
O = os.path.join(os.path.dirname(__file__), "..")
|
||||
old = [json.loads(l) for l in open(os.path.join(O, "work", "existing.jsonl"))]
|
||||
L = lambda s, e, f, d: {"start": s, "end": e, "family": f, "description": d}
|
||||
NEW = {
|
||||
"NORDEND-196": [L("2026-10-16T19:00Z", "2026-10-16T23:00Z", "Visibility",
|
||||
"In early night, the error_value of the reference CTR04 rises and its spot weakens (20:00 image, without visible obstacle): from 19:15 the targets already shift together a little on corr_x (PR16, PR15, PR06, PR08, PR11), then CTR04 is rejected for about an hour: the compensation runs on two references, and almost all the targets jump strongly together on corr_x and corr_y (PR16, PR08, PR15, PR06, PR14 the most), without a step in the raw positions. When CTR04 is found again, its error_value stays raised for more than an hour and the targets keep jumping together on corr_x, one way then the other (PR08, PR12, PR15, PR16, PR06), until CTR04 is back at its level; everything has come back at 23:00. Cause of the weakening not established. The jumps are not a movement; data not to be trusted over the range.")],
|
||||
"NORDEND-209": [L("2026-10-21T10:15Z", "2026-10-21T12:15Z", "Visibility",
|
||||
"In late morning, under an overcast sky, the light dims: on the 10:45 thumbnails the spots are smaller and weaker than at 10:00, nothing in front of the targets. PR09 and PR07 are rejected for a few measurements and their curves are missing (CTR01 and PR04 have their own Target issue labels); meanwhile the corr_x of PR15 drops clearly and those of PR14 and PR16 rise clearly, PR12 and PR13 jump a little, then everything comes back around midday. The references do not move. Cause deduced from the images: the dimmer daylight on these faint targets. Not a movement; ignore the points of the range."),
|
||||
L("2026-10-21T14:15Z", "2026-10-21T17:15Z", "Visibility",
|
||||
"In the afternoon and at dusk, PR09 is rejected without interruption and PR07 at times, their curves missing; the corr_x of PR07 jumps just before. At the change from daylight to night (16:45 image, dark and noisy), the error_value of all the sensors rises for one measurement and PR15 jumps strongly on corr_x and corr_y, PR16 and PR14 clearly, then they come back. The references do not move. Faint targets lost while the light is missing, deduced from the images (overcast sky, nothing in front of them). Not a movement; ignore the points of the range.")],
|
||||
"NORDEND-211": [L("2026-10-22T09:15Z", "2026-10-22T11:00Z", "Visibility",
|
||||
"In mid-morning, under an overcast sky, the light weakens over the whole scene at 09:30: on the thumbnails the spots are smaller and that of PR09 disappears, nothing in front. PR09 is rejected until late morning and PR07 for two measurements, their curves missing; at 09:30 the error_value of PR03, PR05, PR06 and PR15 rises, PR05 jumps strongly on corr_x and corr_y, PR14, PR15 and PR06 clearly, and PR14 stays shifted up on corr_x until 10:15. The references do not move. Cause deduced from the images: the dimmer daylight on faint targets. Not a movement; ignore the points of the range."),
|
||||
L("2026-10-22T12:00Z", "2026-10-22T14:00Z", "Visibility",
|
||||
"Around midday, the light dims again: on the 12:15 and 12:30 thumbnails the spots of PR07, PR09 and PR16 are much weaker than at 11:45 and 14:15, nothing in front. PR09 is rejected for a few measurements, its curve missing; the corr_x of PR07 jumps one way then the other, and PR16, PR15 and PR14 jump clearly for a measurement or two, then everything comes back. The references do not move. Cause deduced from the images: the dimmer daylight on faint targets. Not a movement; ignore the points of the range.")],
|
||||
"NORDEND-212": [L("2026-10-22T20:15Z", "2026-10-23T08:00Z", "Visibility",
|
||||
"PR09, CTR01 and PR04 rejected all night, PR07 at times: at night, the spots of targets on the left edge of the group fade out on the images, without visible obstacle or veil over the rest of the scene; their measurements are rejected and their curves are missing. The other targets and the references remain measured, the camera pose does not move more than usual. At dawn on the 23rd, the first rain veils the whole scene (separate label): PR09 stays rejected until it ends, and the rejections of PR01, PR02, PR03, PR11, PR12 and PR13 at that hour come from it. Cause not established (poorer view at night of these targets, moisture or mist possible). This is not a movement; the curves are missing.")],
|
||||
}
|
||||
ADDED = [
|
||||
L("2026-10-13T12:15Z", "2026-10-13T12:45Z", "Visibility",
|
||||
"Around midday, for one measurement, the spots shrink and weaken over the group (12:30 thumbnails of PR07, PR14 and PR16, against 12:15 and 12:45), under an overcast sky, with nothing in front: PR07 is rejected and its curve is missing, and the corr_x of most of the targets drops clearly together (PR16, PR15, PR14, PR13, PR05, PR06, PR08 the most), the references not moving; everything comes back at the next measurement. Cause seen on the thumbnails: a passing drop of the light on the scene. Not a movement; ignore this point."),
|
||||
L("2026-10-20T15:15Z", "2026-10-20T17:15Z", "Visibility",
|
||||
"In late afternoon, at the change from daylight to night (thumbnails of 15:30, 15:45 and 16:45: the image darkens and becomes noisy), the faint spot of PR07 is rejected for one measurement, its corr_x jumping strongly just before, then jumps again under its threshold until 17:00; PR09 and PR15 jump a little on corr_y at 15:45. The references do not move; PR04, rejected at the same time, has its own Target issue label. Faint target lost while the light is missing, deduced from the images, nothing in front of it. Not a movement; ignore the points of the range."),
|
||||
L("2026-10-20T23:45Z", "2026-10-21T01:45Z", "Visibility",
|
||||
"Around midnight, the spots of PR14, PR15 and PR16 shrink and weaken strongly (00:15 and 00:30 thumbnails, against 23:30 and 01:15), PR12 a little, without a visible obstacle; their error_value rises and PR14 is rejected for one measurement, its curve missing. The corr_y of PR15 and PR16 drop clearly, most targets shift up a little together on corr_x, and the corr_x of PR15 stays lower until 01:30, then everything comes back. The references hold. Cause not established (a local veil at night in front of this part of the group, moisture possible). Not a movement; ignore the points of the range."),
|
||||
L("2026-10-22T08:30Z", "2026-10-22T09:00Z", "Displacement",
|
||||
"For one measurement, PR02 and PR03, neighbours on the wall, jump on corr_x and corr_y (PR03 clearly upward on corr_y), their error_value rising, then come back at the next measurement; the other targets and the references do not move. On the 08:45 thumbnails nothing is seen in front of them. Cause not established (a passage too brief or too close to be seen is possible). A sudden jump, back to normal at once: points to ignore."),
|
||||
L("2026-10-23T05:45Z", "2026-10-23T07:45Z", "Visibility",
|
||||
"At dawn, the first rain of the day: on the 06:45 image the scene is dark and hazy and blurred drops are seen on the window in the lower left. The error_value of all the sensors rises at 06:15; PR01, PR02, PR03, PR11, PR12, PR13 and PR09 are rejected at times and their curves are missing; the corr_x of PR15 and PR16 jump up clearly, those of PR06, PR08 and PR15 down, and at 07:15 most targets are a little low together, then everything comes back. The references stay accepted. Cause seen on the image: rain on the window at dawn. Not a movement; do not trust the curves over the range."),
|
||||
L("2026-10-23T14:15Z", "2026-10-23T14:45Z", "Displacement",
|
||||
"During the rainy afternoon (separate label), the base changes: the references CTR02 and CTR03 move apart on corr_x, and slightly on corr_y with CTR04, without any reference being lost, and from 14:30 all the targets keep a small common shift: the corr_x of PR02, PR03, PR12 and PR01 a little higher, the corr_y of PR06, PR08, PR14, PR15 and PR16 a little lower. The targets do not jump in the raw image. The shift stays after the rain and all through the 24th, without return until the end of the range. Cause not established (water on a reference or on the window in front of it possible). It is not a movement of the structure; to check with the operator if it does not come back."),
|
||||
L("2026-10-23T14:45Z", "2026-10-24T07:30Z", "Thermal displacement",
|
||||
"Cooling of the housing in the evening and the night of the rain: the housing temperature, flat during the rainy day, drops markedly from the afternoon to the middle of the night and stays low until the morning; the camera turns with it, the tilt going down and the pan up, beyond their levels of the previous night, and stays there until the morning; the warming of the 24th brings it back (separate label). The compensation absorbs it: the corrected curves keep nothing of it beyond the jumps of the rain and the base step of 14:30 (separate labels). Deduced cause: the pose follows the housing temperature, and the rock face of the quarry does not move with the temperature. The structure does not move, nothing to do."),
|
||||
L("2026-10-24T14:00Z", "2026-10-24T14:30Z", "Displacement",
|
||||
"For one measurement, the whole image shifts: the raw positions of all the sensors, references included, jump together, the tilt of the camera jumps and the error_value of all the sensors rises; the compensation leaves a small remainder: the corr_y of PR06, PR14, PR15, PR18 and PR13 drop a little, the corr_x of PR16 and PR14 rise a little, then everything comes back. Nothing in front of the targets on the 14:15 thumbnails, overcast sky. Cause not established (a brief shake of the camera possible). A sudden jump, back to normal at once: points to ignore."),
|
||||
L("2026-10-24T15:00Z", "2026-10-24T15:30Z", "Displacement",
|
||||
"For one measurement, without a rise of the error_value except a little on the reference CTR04, the references CTR02 and CTR03 move apart slightly: the base changes, and most targets jump together, up on corr_y (PR16, PR15, PR06, PR13, PR12, PR18 the most) and down on corr_x (PR06, PR11, PR16), then everything comes back at the next measurement. The raw positions of the targets do not jump; the light changes between the 15:00 and 15:15 images, nothing in front of the references. Cause not established. It is not a movement of the structure: points to ignore."),
|
||||
]
|
||||
out = []
|
||||
for d in old:
|
||||
if d["id"] in NEW:
|
||||
out += NEW[d["id"]]
|
||||
else:
|
||||
out.append({k: v for k, v in d.items() if k != "id"})
|
||||
out += ADDED
|
||||
out.sort(key=lambda d: (d["start"], d["end"]))
|
||||
with open(os.path.join(O, "labels.jsonl"), "w") as f:
|
||||
for d in out:
|
||||
f.write(json.dumps(d, ensure_ascii=False) + "\n")
|
||||
print(len(out), "labels;", len(old) - len(NEW), "kept,", len(NEW), "rewritten,", len(ADDED), "added (+ split halves)")
|
||||
Reference in new issue
Block a user