Spaces:
Running
Running
File size: 4,084 Bytes
d70361b | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 | """Re-run detection on drone GT packs and save overlays + summary (report redo).
Library GeoTIFFs for reports 54–59 were not present locally; this uses the
ingested labeling packs (same scenes as before3/4/5/6/1). before7 has no pack.
"""
from __future__ import annotations
import json
import sys
import time
from pathlib import Path
import numpy as np
from PIL import Image
ROOT = Path(__file__).resolve().parent.parent
sys.path.insert(0, str(ROOT))
try:
from dotenv import load_dotenv
load_dotenv(ROOT / ".env")
except ImportError:
pass
PACKS = [
("report_55_style", "dda_before1_after"),
("report_54_style", "dda_before3_after3"),
("report_56_style", "dda_before4_after4"),
("report_57_style", "dda_before5_after5"),
("report_58_style", "dda_before6_after6"),
]
def main():
from app.detection_engine import run_detection
out = ROOT / "data/delhi_cd/friday_drone_report_fix/rerun_overlays"
out.mkdir(parents=True, exist_ok=True)
rows = []
for tag, pid in PACKS:
pack = ROOT / "docs/delhi_eval/dda_labeling" / pid
before = Image.open(pack / "before.png").convert("RGB")
after = Image.open(pack / "after.png").convert("RGB")
if after.size != before.size:
after = after.resize(before.size, Image.Resampling.LANCZOS)
gt_p = ROOT / "docs/delhi_eval/labels" / f"{pid}.png"
t0 = time.time()
mask, vis, stats, regions = run_detection(
before, after,
method="AI-Based Deep Learning",
enable_registration=True,
enable_normalization=True,
detection_sensitivity=0.5,
min_region_area=150,
max_size=max(before.size),
)
elapsed = time.time() - t0
Image.fromarray(vis).save(out / f"{pid}_overlay.png")
pred = np.asarray(mask)
if pred.ndim == 3:
pred = pred[..., 0]
pred = pred > 127
gt = np.array(Image.open(gt_p).convert("L")) > 127
if pred.shape != gt.shape:
gt = np.array(Image.fromarray(gt.astype(np.uint8) * 255).resize(
(pred.shape[1], pred.shape[0]), Image.Resampling.NEAREST)) > 0
p, g = pred.ravel(), gt.ravel()
tp = int(np.logical_and(p, g).sum())
fp = int(np.logical_and(p, ~g).sum())
fn = int(np.logical_and(~p, g).sum())
prec = tp / (tp + fp) if tp + fp else 0.0
rec = tp / (tp + fn) if tp + fn else 0.0
f1 = (2 * prec * rec / (prec + rec)) if prec + rec else 0.0
row = {
"tag": tag,
"pair_id": pid,
"elapsed_s": round(elapsed, 2),
"change_pct": round(float(stats.get("change_percentage") or 0), 3),
"gt_pct": round(100 * float(g.mean()), 3),
"n_regions": len(regions or []),
"f1": round(f1, 4),
"precision": round(prec, 4),
"recall": round(rec, 4),
"registration": (stats.get("params") or {}).get("registration"),
"threshold_debug": {
k: stats.get("threshold_debug", {}).get(k)
for k in (
"pair_ncc", "drone_fp_mode", "drone_tta",
"drone_overfire_clean", "threshold_score",
)
},
"overlay": str((out / f"{pid}_overlay.png").relative_to(ROOT)).replace("\\", "/"),
}
rows.append(row)
print(
f"{pid}: F1={row['f1']:.3f} change%={row['change_pct']:.2f} "
f"gt%={row['gt_pct']:.2f} regions={row['n_regions']}",
flush=True,
)
summary = {
"created_unix": time.time(),
"note": "before7 has no GT pack; library TIFs for reports 54-59 not on disk",
"mean_f1": round(float(np.mean([r["f1"] for r in rows])), 4),
"pairs": rows,
}
(out.parent / "rerun_summary.json").write_text(
json.dumps(summary, indent=2), encoding="utf-8")
print("mean_f1", summary["mean_f1"], "->", out.parent / "rerun_summary.json")
if __name__ == "__main__":
main()
|