How to generate point solutions?

[ImmortalSdm]

Can u give an example code of generating k largest inscribed circles within the mask as described in the implementation details?
Thx a lot~

[LiuRicky]

Thanks for your interest.

We consider release relative codes in a few weeks.

[ImmortalSdm]

Thanks for your interest.

We consider release relative codes in a few weeks.

Glad to hear that! Looking forward to the update.

[LiuRicky]

Thanks for your interest.
We consider release relative codes in a few weeks.

Glad to hear that! Looking forward to the update.

You can try it.

from scipy.ndimage import distance_transform_edt

def get_two_representative_points(m):
    """
    找到两个能较好描述mask形状的点
    
    Args:
        m: 二值图像数组
    
    Returns:
        tuple: ((x1, y1), (x2, y2)) 两个代表性点的坐标
    """
    y_indices, x_indices = np.where(m == 1)
    if len(x_indices) == 0 or len(y_indices) == 0:
        return None, None
    
    # 计算距离变换
    dist_transform = distance_transform_edt(m)
    
    # 找到第一个点（全局最大值点）
    y1, x1 = np.unravel_index(dist_transform.argmax(), dist_transform.shape)
    
    # 计算mask的重心
    center_y = int(np.mean(y_indices))
    center_x = int(np.mean(x_indices))
    
    # 将点分为两组：距离第一个点较远的点和较近的点
    points = np.column_stack((y_indices, x_indices))
    distances_to_first = ((points[:, 0] - y1) ** 2 + (points[:, 1] - x1) ** 2) ** 0.5
    
    # 找到距离第一个点最远的点集
    far_points = points[distances_to_first > np.median(distances_to_first)]
    
    if len(far_points) > 0:
        # 在远点中找到距离变换值最大的点作为第二个点
        far_dist_values = dist_transform[far_points[:, 0], far_points[:, 1]]
        second_point_idx = np.argmax(far_dist_values)
        y2, x2 = far_points[second_point_idx]
    else:
        # 如果没有合适的远点，使用重心附近的点
        local_region = dist_transform[
            max(0, center_y - 10):min(m.shape[0], center_y + 10),
            max(0, center_x - 10):min(m.shape[1], center_x + 10)
        ]
        local_y, local_x = np.unravel_index(local_region.argmax(), local_region.shape)
        y2 = local_y + max(0, center_y - 10)
        x2 = local_x + max(0, center_x - 10)
    
    # 确保两个点都在mask上
    if m[y1, x1] == 0:
        distances = (x_indices - x1)**2 + (y_indices - y1)**2
        nearest_idx = np.argmin(distances)
        x1, y1 = int(x_indices[nearest_idx]), int(y_indices[nearest_idx])
    
    if m[y2, x2] == 0:
        distances = (x_indices - x2)**2 + (y_indices - y2)**2
        nearest_idx = np.argmin(distances)
        x2, y2 = int(x_indices[nearest_idx]), int(y_indices[nearest_idx])
    
    return [x1, y1], [x2, y2]