utility_functions.md

Built-in Utility Functions

In algo/lidar.py (or in configuration under proc/lidar):

• crop: Crops a point cloud
• project_image_pixel_colors: Projects pixel colors from an image to a point cloud, requires calib and image data

In algo/camera.py (or in configuration under proc/camera):

• project_point_cloud_points: Overlays a point cloud on an image, requires calib and point cloud data

In algo/label.py (or in configuration under proc/label):

• remove_out_of_bound_labels: Removes labels that are out of bound of the crop-bound defined in proc/lidar/crop

In algo/post.py (or in configuration under proc/post):

• create_per_object_pcdet_dataset: Extract point-clouds and corresponding bounding-box for each object in a frame, saves point-clouds in .npy format and labels in OpenPCDet annotation format, requires point-cloud and label data
• create_pcdet_dataset: Saves the processed point-cloud data in .npy format and labels in OpenPCDet format, requires point-cloud and label data

Auxiliary Utility Functions

There are some utility functions that are not processing functions but are used in the processing functions. These functions are defined in algo/utils.py and can be imported in any processing function. The utility functions are:

• gather_point_clouds: Gathers current_point_cloud_numpy in data_dict and stores in an array in data_dict using a key and couunt (as the number of point clouds to gather). An argument global_index_key can be used if need to make sure that only previously ungathered point clouds are gathered.
• combine_gathers: Combines gathered point clouds in data_dict using a global_index_key and stores them as single array in data_dict at index key.
• skip_frames: Skips skip number of frames in a sequence of frames. An argument global_index_key can be used if need to make sure that only previously unskipped frames are skipped.