Class DataParser

utils/data_parser.py:124–742

A class for parsing data files in the SceneFun3D dataset.

    return angle_axis.flatten()  # Return as a 1D array (rotation vector)

class DataParser:
    """
    A class for parsing data files in the SceneFun3D dataset.
    """

    def __init__(self, data_root_path):
        """
        Initialize the DataParser instance with the root path.

        Args:
            data_root_path (str): The root path where data is located.
        """
        self.data_root_path = os.path.join(data_root_path)

    def TrajStringToMatrix(self, traj_str):
        """ 
        Converts a line from the camera trajectory file into translation and rotation matrices.

        Args:
            traj_str (str): A space-delimited string where each line represents a camera pose at a particular timestamp. 
                            The line consists of seven columns:
                - Column 1: timestamp
                - Columns 2-4: rotation (axis-angle representation in radians)
                - Columns 5-7: translation (in meters)

        Returns:
            (tuple): A tuple containing:
                - ts (str): Timestamp.
                - Rt (numpy.ndarray): 4x4 transformation matrix representing rotation and translation.

        Raises:
            AssertionError: If the input string does not have exactly seven columns.
        """
        tokens = traj_str.split()
        assert len(tokens) == 7
        ts = tokens[0]

        # Rotation in angle axis
        angle_axis = [float(tokens[1]), float(tokens[2]), float(tokens[3])]
        r_w_to_p = convert_angle_axis_to_matrix3(np.asarray(angle_axis))

        # Translation
        t_w_to_p = np.asarray([float(tokens[4]), float(tokens[5]), float(tokens[6])])
        extrinsics = np.eye(4, 4)
        extrinsics[:3, :3] = r_w_to_p
        extrinsics[:3, -1] = t_w_to_p
        Rt = np.linalg.inv(extrinsics)

        return (ts, Rt)

    def get_camera_trajectory(self, visit_id, video_id, pose_source="colmap"):
        """
        Retrieve the camera trajectory from a file and convert it into a dictionary whose keys are timestamps and 
        values are the corresponding camera poses.

        Args:
            visit_id (str): The identifier of the scene.
            video_id (str): The identifier of the video sequence.