tests/test_evaluate_depth_dataset.py

import json

import numpy as np

from evaluate_depth_dataset import evaluate_manifest


def test_depth_manifest_runner_aggregates_rows_and_records_failures(tmp_path):
    prediction = tmp_path / "prediction.npy"
    ground_truth = tmp_path / "ground_truth.npy"
    np.save(prediction, np.array([[1.0, 2.0], [3.0, 4.0]]))
    np.save(ground_truth, np.array([[1.0, 2.0], [3.0, 4.0]]))
    manifest = tmp_path / "pairs.jsonl"
    manifest.write_text("\n".join([
        json.dumps({"id": "good", "prediction": prediction.name, "ground_truth": ground_truth.name}),
        json.dumps({"id": "missing", "prediction": "missing.npy", "ground_truth": ground_truth.name}),
    ]), encoding="utf-8")
    report = evaluate_manifest(manifest)
    assert report["successful_rows"] == 1
    assert len(report["failed_rows"]) == 1
    assert report["aggregate_median_scaled"]["rmse"]["mean"] == 0.0