utils

General functions for experiments and pytorch.

dict_to(data_dict, device)

Move values in dictionary of type torch.Tensor to a specfied device.

Parameters:

Name	Type	Description	Default
data_dict	dict	Dictionary to be iterated over.	required
device	device	Device to move objects of type torch.Tensor to.	required

Returns: Dictionary containing the same keys and values as data_dict, but with all objects of type torch.Tensor moved to the specified device.

Source code in sdfest/initialization/utils.py

def dict_to(data_dict: dict, device: torch.device) -> dict:
    """Move values in dictionary of type torch.Tensor to a specfied device.

    Args:
        data_dict: Dictionary to be iterated over.
        device: Device to move objects of type torch.Tensor to.
    Returns:
        Dictionary containing the same keys and values as data_dict, but with all
        objects of type torch.Tensor moved to the specified device.
    """
    new_data_dict = {}
    for k, v in data_dict.items():
        if isinstance(v, torch.Tensor):
            new_data_dict[k] = v.to(device)
        else:
            new_data_dict[k] = v
    return new_data_dict

set_axes_equal(ax)

Make axes of 3D plot have equal scale.

This ensures that spheres appear as spheres, cubes as cubes, ... This is needed since Matplotlib's ax.set_aspect('equal') and and ax.axis('equal') are not supported for 3D.

From: https://stackoverflow.com/a/31364297

Parameters:

Name	Type	Description	Default
ax		A Matplotlib axis, e.g., as output from plt.gca().	required

Source code in sdfest/initialization/utils.py

def set_axes_equal(ax) -> None:
    """Make axes of 3D plot have equal scale.

    This ensures that spheres appear as spheres, cubes as cubes, ...
    This is needed since Matplotlib's ax.set_aspect('equal') and
    and ax.axis('equal') are not supported for 3D.

    From: https://stackoverflow.com/a/31364297

    Args:
      ax: A Matplotlib axis, e.g., as output from plt.gca().
    """
    x_limits = ax.get_xlim3d()
    y_limits = ax.get_ylim3d()
    z_limits = ax.get_zlim3d()

    x_range = abs(x_limits[1] - x_limits[0])
    x_middle = np.mean(x_limits)
    y_range = abs(y_limits[1] - y_limits[0])
    y_middle = np.mean(y_limits)
    z_range = abs(z_limits[1] - z_limits[0])
    z_middle = np.mean(z_limits)

    # The plot bounding box is a sphere in the sense of the infinity
    # norm, hence I call half the max range the plot radius.
    plot_radius = 0.5 * max([x_range, y_range, z_range])

    ax.set_xlim3d([x_middle - plot_radius, x_middle + plot_radius])
    ax.set_ylim3d([y_middle - plot_radius, y_middle + plot_radius])
    ax.set_zlim3d([z_middle - plot_radius, z_middle + plot_radius])

str_to_object(name)

Try to find object with a given name.

First scope of calling function is checked for the name, then current environment (in which case name has to be a fully qualified name). In the second case, the object is imported if found.

Parameters:

Name	Type	Description	Default
name	str	Name of the object to resolve.	required

Returns: The object which the provided name refers to. None if no object was found.

Source code in sdfest/initialization/utils.py

def str_to_object(name: str) -> Any:
    """Try to find object with a given name.

    First scope of calling function is checked for the name, then current environment
    (in which case name has to be a fully qualified name). In the second case, the
    object is imported if found.

    Args:
        name: Name of the object to resolve.
    Returns:
        The object which the provided name refers to. None if no object was found.
    """
    # check callers local variables
    caller_locals = inspect.currentframe().f_back.f_locals
    if name in caller_locals:
        return caller_locals[name]

    # check callers global variables (i.e., imported modules etc.)
    caller_globals = inspect.currentframe().f_back.f_globals
    if name in caller_globals:
        return caller_globals[name]

    # check environment
    return locate(name)

visualize_sample(sample=None, prediction=None)

Visualize sample and prediction.

Assumes the following conventions and keys "scale": Half maximum side length of bounding box. "quaternion: Scalar-last orientation of object.

Source code in sdfest/initialization/utils.py

def visualize_sample(sample: Optional[dict] = None, prediction: Optional[dict] = None):
    """Visualize sample and prediction.

    Assumes the following conventions and keys
        "scale": Half maximum side length of bounding box.
        "quaternion: Scalar-last orientation of object.
    """
    pointset = sample["pointset"].cpu().numpy()
    plt.imshow(sample["mask"].cpu().numpy())
    plt.show()
    plt.imshow(sample["depth"].cpu().numpy())
    plt.show()
    fig = plt.figure()
    ax = fig.add_subplot(projection="3d")
    ax.set_box_aspect((1, 1, 1))
    max_points = 500
    if len(pointset) > max_points:
        indices = np.random.choice(len(pointset), replace=False, size=max_points)
        ax.scatter(pointset[indices, 0], pointset[indices, 1], pointset[indices, 2])
    else:
        ax.scatter(pointset[:, 0], pointset[:, 1], pointset[:, 2])

    _plot_coordinate_frame(ax, sample)

    _plot_bounding_box(ax, sample)

    set_axes_equal(ax)

    plt.show()