Create scorer.py

scorer.py

@@ -0,0 +1,65 @@
+import sys
+import pandas as pd
+from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score, confusion_matrix
+
+
+def resolve_label(df: pd.DataFrame) -> pd.Series:
+    label_cols = [c for c in df.columns if c.startswith("label_")]
+    if len(label_cols) == 1:
+        return df[label_cols[0]]
+    raise ValueError(f"Expected one label column, found: {label_cols}")
+
+
+def align_frames(preds: pd.DataFrame, truth: pd.DataFrame) -> tuple[pd.DataFrame, pd.DataFrame]:
+    if len(preds) != len(truth):
+        raise ValueError(
+            f"Row count mismatch: predictions has {len(preds)} rows, truth has {len(truth)} rows"
+        )
+
+    if "scenario_id" in preds.columns and "scenario_id" in truth.columns:
+        preds = preds.sort_values("scenario_id").reset_index(drop=True)
+        truth = truth.sort_values("scenario_id").reset_index(drop=True)
+
+        if not preds["scenario_id"].equals(truth["scenario_id"]):
+            raise ValueError("scenario_id mismatch after alignment between predictions and truth")
+
+    return preds, truth
+
+
+def score(predictions_path: str, ground_truth_path: str) -> dict:
+    preds = pd.read_csv(predictions_path)
+    truth = pd.read_csv(ground_truth_path)
+
+    preds, truth = align_frames(preds, truth)
+
+    y_pred = resolve_label(preds)
+    y_true = resolve_label(truth)
+
+    accuracy = accuracy_score(y_true, y_pred)
+    precision = precision_score(y_true, y_pred, zero_division=0)
+    recall = recall_score(y_true, y_pred, zero_division=0)
+    f1 = f1_score(y_true, y_pred, zero_division=0)
+
+    tn, fp, fn, tp = confusion_matrix(y_true, y_pred, labels=[0, 1]).ravel()
+
+    false_safe_rate = fn / (fn + tp) if (fn + tp) > 0 else 0.0
+
+    return {
+        "accuracy": accuracy,
+        "precision": precision,
+        "recall_cascade_detection": recall,
+        "false_safe_rate": false_safe_rate,
+        "f1": f1,
+        "confusion_matrix": {
+            "tp": int(tp),
+            "fp": int(fp),
+            "tn": int(tn),
+            "fn": int(fn),
+        },
+    }
+
+
+if __name__ == "__main__":
+    if len(sys.argv) != 3:
+        raise SystemExit("Usage: python scorer.py <predictions.csv> <ground_truth.csv>")
+    print(score(sys.argv[1], sys.argv[2]))