Did you use the calibrations data?

[magical1998]

    ##################
    # Load calibration
    ##################

    # Init variables
    self.calibrations = []
    self.times = []
    self.poses = []
    self.all_inds = None
    self.class_proportions = None
    self.class_frames = []
    self.val_confs = []

    # Load everything
    self.load_calib_poses()

This code is from semantickitti.py, but I don't seem to see any subsequent information about calibrations, but the poses information is indeed used as an alignment between frames, am I seeing this correctly, or am I missing a step?

[magical1998]

I seem to understand, you seem to have put all the pose information into the camera coordinate system for calibration, where the calibrations information provides a calibration matrix for the poses information.