Preparing Your Data

Do you have your data ready?

Point Cloud Data [Required]

Scale’s platform expects at least one point cloud per frame (the full sweep). If you have multiple lidar sensors, you will want to combine those into a singular point cloud.

The format of these files will vary, common examples include .pcd , .json , .bin .

Lidar Sensor Calibrations [Required]

Scale’s platform expects a calibration file, which is used to calculate your device’s heading.

You will also need to provide an “ego2world” or “poses” file which is used to calculate your device’s location in the point cloud.

It is strongly advised that your data is submitted in a static / world-frame of reference such that the ego device moves throughout the scene, as opposed to an “ego-only” coordinate system, in which the device remains centered and the scene moves around the device.

Camera Images and Associated Camera Calibrations [Optional, but very helpful]

Scale’s platform expects images to be in a browser-viewable format such as jpg or png .

For each camera, you will need to provide the extrinsic and intrinsic calibration matrices, which are used for features such as cuboid projections and point projections (see reference docs, “Camera Image” object. )

Other Calibration and Transformation Data [Optional]

You may also include calibration data such as lidar2Cam , world2Lidar , etc. You will need to adjust the script to include and use them.