utils module

utils.compute_ate(gtruth_xyz, pred_xyz_o)
utils.compute_errors(gt, pred)

Computation of error metrics (abs rel,sq rel, rmse, rmse log) between predicted and ground truth depths From https://github.com/mrharicot/monodepth

Parameters

  • gt – an array with the ground truth values

  • pred – an array with the predicted values

Returns

the error

Return type

abs_rel,sq_rel,rmse,rmse_log

utils.convert_scale(points, gt_depth)

convert the scale of the predictions to the gt

Parameters

  • points – the predictions depth

  • gt_filename – the gt references filename

Returns: the ratio between the predictions and the gt

utils.create_dir(directory)

Create a directory if not exists :param directory: directory to create

Returns

None, but it creates a new folder if not exists

utils.dump_xyz(source_to_target_transformations)
utils.evaluate_pose(args)

Evaluate odometry on the KITTI dataset

utils.get_error(args, filename, points, gt_filename)

Get the realtive gt from it’s filename and convert the scale of the predictions in order to compute the error

Parameters

  • points – the predictions depth

  • gt_filename – the gt references filename

Returns

the error computed on this examples

utils.get_error_KITTI(points, gt_filename)

Get the realtive gt from it’s filename and convert the scale of the predictions in order to compute the error

Parameters

  • points – the predictions depth

  • gt_filename – the gt references filename

Returns

the error computed on this examples

utils.get_error_TUM(points, gt_filename)

Get the realtive gt from it’s filename and convert the scale of the predictions in order to compute the error

Parameters

  • points – the predictions depth

  • gt_filename – the gt references filename

Returns

the error computed on this examples

utils.load_IMU_datas_TUM_VI(path_to_sequence)
utils.load_images_EuRoC(path_to_sequence)

This loader is created for Visual Inertial EuRoC datasets. Format of such datasets is: path_to_sequence/mav0/cam0/+data/xxxx.png

/-times.txt

utils.load_images_KITTI(path_to_sequence)

Return the sequence of the images found in the path and the corrispondent timestamp

Parameters

path_to_sequence – the sequence in witch we can found the image sequences

Returns :

two array : one contains the sequence of the image filename and the second the timestamp in whitch they are acquired

utils.load_images_KITTI_VO(path_to_sequence)

Return the sequence of the images found in the path and the corrispondent timestamp

Parameters

path_to_sequence – the sequence in witch we can found the image sequences

Returns :

two array : one contains the sequence of the image filename and the second the timestamp in whitch they are acquired

utils.load_images_OTHERS(path_to_sequence)

Return the sequence of the images found in the path and the corrispondent timestamp

Parameters

path_to_sequence – the sequence in witch we can found the image sequences

Returns :

two array : one contains the sequence of the image filename and the second the timestamp in whitch they are acquired

Inside of path_to_sequence must be: +data

xxxxxxxx.png xxxxxxxy.png ….

-times.txt

where times.txt simply contains timestamps of every frame

utils.load_images_TUM(path_to_sequence, file_name)

Return the sequence of the images found in the path and the corrispondent timestamp

Parameters

path_to_sequence – the sequence in witch we can found the image sequences

Returns

one contains the sequence of the image filename and the second the timestamp in whitch they are acquired

Return type

two array

utils.load_images_TUM_VI(path_to_sequence)

This loader is created for Visual Inertial TUM datasets. Format of such datasets is: path_to_sequence/mav0/cam0/+data/xxxx.png

/-times.txt

utils.read_depth_KITTI(filename)

loads depth map D from png file and returns it as a numpy array,

utils.read_depth_TUM(filename)

loads depth map D from png file and returns it as a numpy array,

utils.save_depth(dest, depth)

Save depth as 16 bit png file

Parameters

  • dest – path to new 16 bit png image wiht depth, w/o exension

  • depth – depth to save, as ndarray HxW

Returns

None, but a new 16 bit png image will be saved at dest

utils.save_depth_err_results(file_path, filename, err)
utils.save_pose(dest, pose)

Save pose as npy file

Parameters

  • dest – path to new npy file wiht pose, w/o exension

  • pose – ndarray with 4x4 pose matrix (as R|t in homogeneous notation)

Returns

None, but it creates a new npy file with the pose

utils.save_pose_and_times_txt(args, name, pose)

Save pose and time in two different txt files.

utils.save_pose_txt(args, name, pose)

Save pose as txt file

Parameters

  • dir – directory of pose.txt

  • name – frame name or id

  • pose – ndarray with 4x4 pose matrix (as R|t in homogeneous notation)

Returns

None, but it creates a new npy file with the pose