def compute_metrics_track2()

in scripts/compute_metrics.py [0:0]

• 46 lines of code
• 8 McCabe index (conditional complexity)

def compute_metrics_track2(args):
    gt = read_ground_truth(args.gt_filepath)



    db_image_ids, xb = read_descriptors(args.db_descs)
    d = xb.shape[1]

    if d > args.max_dim:
        raise RuntimeError(
            f"maximum dimension exceeded {d} > {args.max_dim}"
        )

    query_image_ids, xq = read_descriptors(args.query_descs)

    if d != xq.shape[1]:
        raise RuntimeError(
            f"query descriptors ({xq.shape[1]}) not same dimension as db ({d})"
        )

    if args.knn == -1:
        print(
            f"Track 2 running matching of {len(query_image_ids)} queries in "
            f"{len(db_image_ids)} database ({d}D descriptors), "
            f"max_results={args.max_results}."
        )
        predictions = match_and_make_predictions(
            xq, query_image_ids,
            xb, db_image_ids,
            args.max_results,
            metric = faiss.METRIC_INNER_PRODUCT if args.ip else faiss.METRIC_L2
        )
    else:
        print(
            f"Track 2 running matching of {len(query_image_ids)} queries in "
            f"{len(db_image_ids)} database ({d}D descriptors), "
            f"kNN with k={args.knn}."
        )
        predictions = knn_match_and_make_predictions(
            xq, query_image_ids,
            xb, db_image_ids,
            args.knn,
            metric = faiss.METRIC_INNER_PRODUCT if args.ip else faiss.METRIC_L2
        )

    if args.write_predictions:
        print("writing predictions to", args.write_predictions)
        write_predictions(predictions, args.write_predictions)

    print(f"Evaluating {len(predictions)} predictions ({len(gt)} GT matches)")

    metrics = evaluate(gt, predictions)
    print_metrics(metrics)
    if args.pr_curve_filepath:
        plot_pr_curve(metrics, args.title, args.pr_curve_filepath)
    return metrics