I need to create AI which transform pixilated image to vector. The input data are photos of manually drawn flowcharts. Their vector representations have been created. I plan to use TensorFlow. As I understand it, vector images need to be converted to a numpy array, but I can’t find a way to convert. How can this problem be solved?

for element in path_elements:
    d = element.get("d")
    path_data = d.split()
    prev_x = None
    prev_y = None
    print(type(path_data[0]))
    for i, segment in enumerate(path_data):
        if segment.startswith("M") or segment.startswith("L"):
            coords = path_data[i+1].split(",")
            print(coords)
            if len(coords) == 2:
                x, y = map(float, coords)
                nodes.append((x, y))
                if prev_x is not None and prev_y is not None:
                    edges.append(((prev_x, prev_y), (x, y)))
                prev_x = x
                prev_y = y
    if not nodes:
        print(f"В элементе path '{element.get('id')}' нет корректных координат.")
if not nodes:
    print("Массив nodes пуст, невозможно выполнить нормализацию.")
else:
    nodes_norm = (np.array(nodes) - np.array(nodes).min(axis=0)) / (np.array(nodes).max(axis=0) - np.array(nodes).min(axis=0)
   vector_representation = np.concatenate([nodes_norm.flatten(), np.array(edges).flatten()])

This code extracts only the coordinates of points and lines. However, what about other commands if I need a two-dimensional array?