cksghl1004/cpp_moondream2

1import numpy as np2 3 4def remove_outlier_points(points_tuples, k_nearest=2, threshold=2.0):5    """6    Robust outlier detection for list of (x,y) tuples.7    Only requires numpy.8 9    Args:10        points_tuples: list of (x,y) tuples11        k_nearest: number of neighbors to consider12        threshold: multiplier for median distance13 14    Returns:15        list: filtered list of (x,y) tuples with outliers removed16        list: list of booleans indicating which points were kept (True = kept)17    """18    points = np.array(points_tuples)19    n_points = len(points)20 21    # Calculate pairwise distances manually22    dist_matrix = np.zeros((n_points, n_points))23    for i in range(n_points):24        for j in range(i + 1, n_points):25            # Euclidean distance between points i and j26            dist = np.sqrt(np.sum((points[i] - points[j]) ** 2))27            dist_matrix[i, j] = dist28            dist_matrix[j, i] = dist29 30    # Get k nearest neighbors' distances31    k = min(k_nearest, n_points - 1)32    neighbor_distances = np.partition(dist_matrix, k, axis=1)[:, :k]33    avg_neighbor_dist = np.mean(neighbor_distances, axis=1)34 35    # Calculate mask using median distance36    median_dist = np.median(avg_neighbor_dist)37    mask = avg_neighbor_dist <= threshold * median_dist38 39    # Return filtered tuples and mask40    filtered_tuples = [t for t, m in zip(points_tuples, mask) if m]41    return filtered_tuples42