File size: 8,064 Bytes
6d5f1ce 5265d5f 6d5f1ce 1262684 6d5f1ce 1262684 6d5f1ce 1262684 6d5f1ce 1262684 6d5f1ce 1262684 6d5f1ce 5265d5f 6d5f1ce 1262684 6d5f1ce 5265d5f 6d5f1ce 5265d5f 6d5f1ce 1262684 5265d5f 6d5f1ce 5265d5f 6d5f1ce 1262684 6d5f1ce | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 | #!/usr/bin/env python3
"""Run EffB2 prediction QC for paired diffusion augmentation and print confidence."""
from __future__ import annotations
import argparse
import csv
import subprocess
import sys
from collections import defaultdict
from pathlib import Path
def parse_args() -> argparse.Namespace:
parser = argparse.ArgumentParser(description="Run EffB2 QC prediction and print confidence summary.")
checkpoint_group = parser.add_mutually_exclusive_group(required=True)
checkpoint_group.add_argument("--checkpoint", type=Path, help="Path to one classifier best.pt checkpoint.")
checkpoint_group.add_argument(
"--checkpoint-dir",
type=Path,
help="Run directory containing fold_*/best.pt; all folds are ensembled for QC.",
)
parser.add_argument("--output-dir", type=Path, default=Path("Stable_diffusion_augmentation/out_minority_pairs"))
parser.add_argument("--batch-size", type=int, default=16)
parser.add_argument("--image-size", type=int, default=384)
parser.add_argument("--num-workers", type=int, default=0)
parser.add_argument("--python", default=sys.executable, help="Python executable to use for prediction.")
parser.add_argument(
"--predict-script",
type=Path,
default=None,
help="Path to predict_milk10k_effb2_dual_metadata.py. Defaults to auto-detect from repo root.",
)
parser.add_argument(
"--summary-script",
type=Path,
default=None,
help="Path to summarize_effb2_qc.py. Defaults to this script's folder.",
)
parser.add_argument("--print-misses", type=int, default=20, help="Number of wrong-target rows to print.")
return parser.parse_args()
def run_command(cmd: list[str]) -> None:
print("Running:")
print(" " + " ".join(cmd))
subprocess.run(cmd, check=True)
def default_repo_root() -> Path:
return Path(__file__).resolve().parents[1]
def resolve_script(path: Path | None, default_path: Path, label: str) -> Path:
script = (path or default_path).expanduser().resolve()
if not script.exists():
raise FileNotFoundError(f"{label} not found: {script}")
return script
def read_rows(path: Path) -> list[dict[str, str]]:
with path.open(newline="") as f:
return list(csv.DictReader(f))
def as_float(value: str) -> float:
try:
return float(value)
except (TypeError, ValueError):
return 0.0
def print_confidence_summary(summary_path: Path, print_misses: int) -> None:
rows = read_rows(summary_path)
if not rows:
print("No QC rows found.")
return
correct = [row for row in rows if row["is_target_predicted"] == "True"]
by_class: dict[str, list[dict[str, str]]] = defaultdict(list)
for row in rows:
by_class[row["target_class"]].append(row)
print("")
print("EffB2 QC confidence summary")
print(f" Total synthetic pairs: {len(rows)}")
print(f" Target predicted: {len(correct)}/{len(rows)} ({len(correct) / len(rows):.1%})")
for class_name in sorted(by_class):
class_rows = by_class[class_name]
class_correct = [row for row in class_rows if row["is_target_predicted"] == "True"]
avg_conf = sum(as_float(row["confidence"]) for row in class_rows) / len(class_rows)
avg_target_prob = sum(as_float(row["target_class_probability"]) for row in class_rows) / len(class_rows)
pred_counts: dict[str, int] = defaultdict(int)
for row in class_rows:
pred_counts[row["label_pred"]] += 1
top_preds = ", ".join(f"{label}:{count}" for label, count in sorted(pred_counts.items(), key=lambda item: (-item[1], item[0]))[:5])
print(
f" {class_name}: target_predicted={len(class_correct)}/{len(class_rows)} "
f"({len(class_correct) / len(class_rows):.1%}), "
f"avg_confidence={avg_conf:.4f}, avg_target_prob={avg_target_prob:.4f}, "
f"top_preds=[{top_preds}]"
)
misses = [row for row in rows if row["is_target_predicted"] != "True"]
misses.sort(key=lambda row: as_float(row["target_class_probability"]))
if misses and print_misses > 0:
print("")
print(f"Lowest target-probability misses, first {min(print_misses, len(misses))}:")
for row in misses[:print_misses]:
print(
f" {row['synthetic_lesion_id']}: target={row['target_class']} "
f"pred={row['label_pred']} conf={as_float(row['confidence']):.4f} "
f"target_prob={as_float(row['target_class_probability']):.4f}"
)
if any(row.get("source_lesion_id") for row in rows):
print("")
print("Worst source lesions by target probability:")
by_source: dict[str, list[dict[str, str]]] = defaultdict(list)
for row in rows:
by_source[row.get("source_lesion_id", "")].append(row)
source_stats = []
for source_lesion_id, source_rows in by_source.items():
avg_target_prob = sum(as_float(row["target_class_probability"]) for row in source_rows) / len(source_rows)
target_predicted = sum(1 for row in source_rows if row["is_target_predicted"] == "True")
source_stats.append((avg_target_prob, source_lesion_id, target_predicted, len(source_rows)))
for avg_target_prob, source_lesion_id, target_predicted, total in sorted(source_stats)[:10]:
print(
f" {source_lesion_id}: target_predicted={target_predicted}/{total} "
f"({target_predicted / total:.1%}), avg_target_prob={avg_target_prob:.4f}"
)
def main() -> None:
args = parse_args()
output_dir = args.output_dir.expanduser().resolve()
checkpoint = args.checkpoint.expanduser().resolve() if args.checkpoint else None
checkpoint_dir = args.checkpoint_dir.expanduser().resolve() if args.checkpoint_dir else None
manifest = output_dir / "paired_augmentation_manifest.csv"
metadata_csv = output_dir / "metadata_for_prediction.csv"
groundtruth_csv = output_dir / "groundtruth_for_prediction.csv"
input_dir = output_dir / "prediction_input"
predictions = output_dir / "effb2_qc_predictions.csv"
summary = output_dir / "effb2_qc_summary.csv"
repo_root = default_repo_root()
predict_script = resolve_script(args.predict_script, repo_root / "predict_milk10k_effb2_dual_metadata.py", "Predict script")
summary_script = resolve_script(
args.summary_script,
Path(__file__).resolve().parent / "summarize_effb2_qc.py",
"Summary script",
)
for path in (checkpoint or checkpoint_dir, manifest, metadata_csv, groundtruth_csv, input_dir):
if not path.exists():
raise FileNotFoundError(f"Required QC input not found: {path}")
predict_command = [
args.python,
str(predict_script),
]
if checkpoint is not None:
predict_command.extend(["--checkpoint", str(checkpoint)])
else:
predict_command.extend(["--checkpoint-dir", str(checkpoint_dir)])
predict_command.extend([
"--input-dir",
str(input_dir),
"--metadata-csv",
str(metadata_csv),
"--groundtruth-csv",
str(groundtruth_csv),
"--output",
str(predictions),
"--include-debug-columns",
"--batch-size",
str(args.batch_size),
"--image-size",
str(args.image_size),
"--num-workers",
str(args.num_workers),
])
run_command(predict_command)
run_command(
[
args.python,
str(summary_script),
"--manifest",
str(manifest),
"--predictions",
str(predictions),
"--output",
str(summary),
]
)
print_confidence_summary(summary, args.print_misses)
print("")
print(f"Predictions CSV: {predictions}")
print(f"QC summary CSV: {summary}")
if __name__ == "__main__":
main()
|