evo_ape tum指令使用及各参数使用效果

目录

参数含义

效果

evo_ape tum gt.txt l.txt

 算法选项

evo_ape tum gt.txt l.txt -a -p 

evo_ape tum gt.txt l.txt -s -p

evo_ape tum gt.txt l.txt -as -p

evo_ape tum gt.txt l.txt  --n_to_align 1

输出选项

evo_ape tum gt.txt l.txt -p

evo_ape tum gt.txt l.txt -as -p --plot_mode=xy

evo_ape tum gt.txt l.txt -as -p  --plot_colormap_max 0.2

evo_ape tum gt.txt l.txt -as -p --plot_colormap_min 0.2

evo_ape tum gt.txt l.txt -as -p --plot_colormap_max_percentile 50

evo_ape tum gt.txt l.txt -as -p  --plot_full_ref

evo_ape tum gt.txt l.txt -as  --save_plot SAVE_PLOT.pdf

evo_ape tum gt.txt l.txt -as --serialize_plot SERIALIZE_PLOT

evo_ape tum gt.txt l.txt -as --save_results ape_result.zip

evo_ape tum gt.txt l.txt -as --logfile ape_result.log

可用性选项

evo_ape tum gt.txt l.txt -as  --no_warnings

evo_ape tum gt.txt l.txt -v

evo_ape tum gt.txt l.txt -as --silent

 evo_ape tum gt.txt l.txt -as   --debug

---

学习evo_ape tum指令的使用

参数含义

查看帮助

evo_ape tum -h

$ evo_ape tum -h
usage: evo_ape tum [-h] [-r {full,trans_part,rot_part,angle_deg,angle_rad}]
                   [-a] [-s] [--n_to_align N_TO_ALIGN] [--align_origin] [-p]
                   [--plot_mode {xy,xz,yx,yz,zx,zy,xyz}]
                   [--plot_colormap_max PLOT_COLORMAP_MAX]
                   [--plot_colormap_min PLOT_COLORMAP_MIN]
                   [--plot_colormap_max_percentile PLOT_COLORMAP_MAX_PERCENTILE]
                   [--plot_full_ref] [--ros_map_yaml ROS_MAP_YAML]
                   [--save_plot SAVE_PLOT] [--serialize_plot SERIALIZE_PLOT]
                   [--save_results SAVE_RESULTS] [--logfile LOGFILE]
                   [--no_warnings] [-v] [--silent] [--debug] [-c CONFIG]
                   [--t_max_diff T_MAX_DIFF] [--t_offset T_OFFSET]
                   ref_file est_file

Absolute pose error (APE) metric app for TUM trajectory files - (c) evo
authors

positional arguments:
  ref_file              reference trajectory file
  est_file              estimated trajectory file

optional arguments:
  -h, --help            show this help message and exit
  --t_max_diff T_MAX_DIFF
                        maximum timestamp difference for data association
  --t_offset T_OFFSET   constant timestamp offset for data association

algorithm options:
  -r {full,trans_part,rot_part,angle_deg,angle_rad}, --pose_relation {full,trans_part,rot_part,angle_deg,angle_rad}
                        pose relation on which the APE is based
  -a, --align           alignment with Umeyama's method (no scale)
  -s, --correct_scale   correct scale with Umeyama's method
  --n_to_align N_TO_ALIGN
                        the number of poses to use for Umeyama alignment,
                        counted from the start (default: all)
  --align_origin        align the trajectory origin to the origin of the
                        reference trajectory

output options:
  -p, --plot            show plot window
  --plot_mode {xy,xz,yx,yz,zx,zy,xyz}
                        the axes for plot projection
  --plot_colormap_max PLOT_COLORMAP_MAX
                        the upper bound used for the color map plot (default:
                        maximum error value)
  --plot_colormap_min PLOT_COLORMAP_MIN
                        the lower bound used for the color map plot (default:
                        minimum error value)
  --plot_colormap_max_percentile PLOT_COLORMAP_MAX_PERCENTILE
                        percentile of the error distribution to be used as the
                        upper bound of the color map plot (in %, overrides
                        --plot_colormap_max)
  --plot_full_ref       plot the full, unsynchronized reference trajectory
  --ros_map_yaml ROS_MAP_YAML
                        yaml file of an ROS 2D map image (.pgm/.png) that will
                        be drawn into the plot
  --save_plot SAVE_PLOT
                        path to save plot
  --serialize_plot SERIALIZE_PLOT
                        path to serialize plot (experimental)
  --save_results SAVE_RESULTS
                        .zip file path to store results
  --logfile LOGFILE     Local logfile path.

usability options:
  --no_warnings         no warnings requiring user confirmation
  -v, --verbose         verbose output
  --silent              don't print any output
  --debug               verbose output with additional debug info
  -c CONFIG, --config CONFIG
                        .json file with parameters (priority over command line
                        args)

翻译之后

$ evo_ape tum -h
usage: evo_ape tum [-h] [-r {full,trans_part,rot_part,angle_deg,angle_rad}]
                   [-a] [-s] [--n_to_align N_TO_ALIGN] [--align_origin] [-p]
                   [--plot_mode {xy,xz,yx,yz,zx,zy,xyz}]
                   [--plot_colormap_max PLOT_COLORMAP_MAX]
                   [--plot_colormap_min PLOT_COLORMAP_MIN]
                   [--plot_colormap_max_percentile PLOT_COLORMAP_MAX_PERCENTILE]
                   [--plot_full_ref] [--ros_map_yaml ROS_MAP_YAML]
                   [--save_plot SAVE_PLOT] [--serialize_plot SERIALIZE_PLOT]
                   [--save_results SAVE_RESULTS] [--logfile LOGFILE]
                   [--no_warnings] [-v] [--silent] [--debug] [-c CONFIG]
                   [--t_max_diff T_MAX_DIFF] [--t_offset T_OFFSET]
                   ref_file est_file

Absolute pose error (APE) metric app for TUM trajectory files - (c) evo
authors
TUM轨迹文件的绝对姿态误差（APE）度量应用程序-（c）evo

作者
positional arguments:
位置参数：
  ref_file              reference trajectory file参考轨迹文件
  est_file              estimated trajectory file估计轨迹文件

optional arguments:可选参数：
  -h, --help            show this help message and exit显示此帮助消息并退出
  --t_max_diff T_MAX_DIFF
                        maximum timestamp difference for data association数据关联的最大时间戳差异
  --t_offset T_OFFSET   constant timestamp offset for data association数据关联的常量时间戳偏移量

algorithm options:
算法选项：
  -r {full,trans_part,rot_part,angle_deg,angle_rad}, --pose_relation {full,trans_part,rot_part,angle_deg,angle_rad}
                           pose relation on which the APE is basedAPE所基于的姿态关系
  -a, --align           alignment with Umeyama's method (no scale)与Umeyama的方法对齐（无比例）
  -s, --correct_scale   correct scale with Umeyama's method用Umeyama的方法校正刻度
  --n_to_align N_TO_ALIGN
                        the number of poses to use for Umeyama alignment,counted from the start (default: all)用于Umeyama对齐的姿势数，从开始计数（默认值：全部）
  --align_origin        align the trajectory origin to the origin of the reference trajectory将轨迹原点与参考轨迹的原点对齐

output options:输出选项：
  -p, --plot            show plot window显示绘图窗口
  --plot_mode {xy,xz,yx,yz,zx,zy,xyz}
                        the axes for plot projection绘图投影的轴
  --plot_colormap_max PLOT_COLORMAP_MAX
                        the upper bound used for the color map plot (default: maximum error value)用于颜色映射图的上限（默认值：最大误差值）
  --plot_colormap_min PLOT_COLORMAP_MIN
                        the lower bound used for the color map plot (default:minimum error value)用于颜色映射图的下限（默认值：最小误差值）
  --plot_colormap_max_percentile PLOT_COLORMAP_MAX_PERCENTILE
                        percentile of the error distribution to be used as the  upper bound of the color map plot (in %, overrides  --plot_colormap_max)要用作颜色映射图上限的错误分布的百分位数（以%为单位，覆盖--plot_colormap_max）
  --plot_full_ref       plot the full, unsynchronized reference trajectory绘制完整的、不同步的参考轨迹
  --ros_map_yaml ROS_MAP_YAML
                        yaml file of an ROS 2D map image (.pgm/.png) that will be drawn into the plot
  --save_plot SAVE_PLOT
                        path to save plot保存绘图的路径
  --serialize_plot SERIALIZE_PLOT
                        path to serialize plot (experimental)序列化绘图的路径（实验）
  --save_results SAVE_RESULTS
                        .zip file path to store results存储结果的.zip文件路径
  --logfile LOGFILE     Local logfile path.本地日志文件路径。

usability options:可用性选项：
  --no_warnings         no warnings requiring user confirmation无需用户确认的警告
  -v, --verbose         verbose output详细输出
  --silent              don't print any output不打印任何输出
  --debug               verbose output with additional debug info使用附加调试信息调试详细输出
  -c CONFIG, --config CONFIG
                        .json file with parameters (priority over command line args)带有参数的.json文件（优先级高于命令行args）

效果

受到数据影响，很多参数的效果无法有效展示。

evo_ape tum gt.txt l.txt

$ evo_ape tum gt.txt l.txt
APE w.r.t. translation part (m)
(not aligned)

       max    6372512.914678
      mean    6372511.473926
    median    6372511.871822
       min    6372508.722174
      rmse    6372511.473926
       sse    934004757162323.000000
       std    1.138090

 算法选项

--t_max_diff T_MAX_DIFF
                        maximum timestamp difference for data association数据关联的最大时间戳差异
  --t_offset T_OFFSET   constant timestamp offset for data association数据关联的常量时间戳偏移量

如果两个txt的实际时间戳存在固定差值m，可以用“  --t_offset  m”

evo_ape tum gt.txt l.txt -a -p 

  -a, --align           alignment with Umeyama's method (no scale)用Umeyama的方法对齐（无比例）

$ evo_apeum gt.txt l.txt -a -p
APE w.r.t. translation part (m)
(with SE(3) Umeyama alignment)

       max    0.869683
      mean    0.252346
    median    0.222082
       min    0.024169
      rmse    0.330618
       sse    2.514097
       std    0.213612

evo_ape tum gt.txt l.txt -s -p

$ evo_ape tum gt.txt l.txt -s -p
APE w.r.t. translation part (m)
(scale corrected)

       max    6372512.113874
      mean    6372510.886484
    median    6372511.224692
       min    6372508.535736
      rmse    6372510.886484
       sse    934004584962312.250000
       std    0.971000

evo_ape tum gt.txt l.txt -as -p

同时用多个参数可以写在一起，比如“-a -s”写作“-as”。

$ evo_ape tum gt.txt l.txt -as -p
APE w.r.t. translation part (m)
(with Sim(3) Umeyama alignment)

       max    0.384438
      mean    0.156007
    median    0.141948
       min    0.010909
      rmse    0.185972
       sse    0.795467
       std    0.101230

You can use Umeyama alignment as a pre-processing step:

    --align or -a = SE(3) Umeyama alignment (rotation, translation)
    --align --correct_scale or -as = Sim(3) Umeyama alignment (rotation, translation, scale)
    --correct_scale or -s = scale alignment

Scale or Sim(3) alignment is usually required for monocular SLAM, where you usually have random scale. SE(3) alignment is useful for the absolute pose error (evo_ape) if you want to measure the shape similarity of trajectories as best as possible. The alignment_demo.py script shows the different types of alignment with an example trajectory (as shown in the evo_traj documentation).

New since v1.5.0: A simple origin alignment that can be useful for drift/loop closure evaluation is available with --align_origin. This is not based on the Umeyama algorithm.

您可以使用Umeyama对齐作为预处理步骤：
--align或-a是SE（3）Umeyama对齐（旋转、平移）
--align—correct_scale或-as 是Sim（3）Umeyama对齐（旋转、平移、缩放）
--correct_scale或-s是缩放对齐
单目SLAM通常需要缩放或Sim（3）对齐，其中通常有随机缩放。如果你想尽可能最好地测量轨迹的形状相似性，SE（3）对齐对绝对姿态误差（evo_ape）很有用。alignment_demo.py脚本显示了具有示例轨迹的不同类型的对齐（如evo_traj文档中所示）。
v1.5.0以来的新增功能：--align_origin提供了一种简单的原点对齐，可用于漂移/环路闭合评估。这不是基于Umeyama算法。

evo_ape tum gt.txt l.txt  --n_to_align 1

  --n_to_align N_TO_ALIGN
                        the number of poses to use for Umeyama alignment,counted from the start (default: all)用于Umeyama对齐的姿势数，从开始计数（默认值：全部）

$ evo_ape tum gt.txt l.txt  --n_to_align 1
APE w.r.t. translation part (m)
(not aligned)

       max    6372512.914678
      mean    6372511.473926
    median    6372511.871822
       min    6372508.722174
      rmse    6372511.473926
       sse    934004757162323.000000
       std    1.138090

输出选项

evo_ape tum gt.txt l.txt -p

  -p, --plot            show plot window显示绘图窗口

$ evo_ape tum gt.txt l.txt -p
APE w.r.t. translation part (m)
(not aligned)

       max    6372512.914678
      mean    6372511.473926
    median    6372511.871822
       min    6372508.722174
      rmse    6372511.473926
       sse    934004757162323.000000
       std    1.138090

evo_ape tum gt.txt l.txt -as -p --plot_mode=xy

--plot_mode {xy,xz,yx,yz,zx,zy,xyz}
                        the axes for plot projection绘图投影的轴

$ evo_ape tum gt.txt l.txt -as -p --plot_mode=xy

绘图投影在xy平面上。

evo_ape tum gt.txt l.txt -as -p  --plot_colormap_max 0.2

  --plot_colormap_max PLOT_COLORMAP_MAX
                        the upper bound used for the color map plot (default: maximum error value)用于颜色映射图的上限（默认值：最大误差值）

$ evo_ape tum gt.txt l.txt -as -p  --plot_colormap_max 0.2

颜色映射图的上限设为0.2，即0.2为红黑色。

evo_ape tum gt.txt l.txt -as -p --plot_colormap_min 0.2

  --plot_colormap_min PLOT_COLORMAP_MIN
                        the lower bound used for the color map plot (default:minimum error value)用于颜色映射图的下限（默认值：最小误差值）

$ evo_ape tum gt.txt l.txt -as -p --plot_colormap_min 0.2

颜色映射图的下限设为0.2，即0.2为蓝黑色。

evo_ape tum gt.txt l.txt -as -p --plot_colormap_max_percentile 50

  --plot_colormap_max_percentile PLOT_COLORMAP_MAX_PERCENTILE
                        percentile of the error distribution to be used as the  upper bound of the color map plot (in %, overrides  --plot_colormap_max)要用作颜色映射图上限的错误分布的百分位数（以%为单位，覆盖--plot_colormap_max）

$ evo_apeum gt.txt l.txt -as -p --plot_colormap_max_percentile 50
APE w.r.t. translation part (m)
(with Sim(3) Umeyama alignment)

       max    0.384438
      mean    0.156007
    median    0.141948
       min    0.010909
      rmse    0.185972
       sse    0.795467
       std    0.101230

evo_ape tum gt.txt l.txt -as -p  --plot_full_ref

  --plot_full_ref       plot the full, unsynchronized reference trajectory绘制完整的、不同步的参考轨迹

$ evo_ape tum gt.txt l.txt -as -p  --plot_full_ref

（略）

我的理解是绘制轨迹没有平滑，存在原始数据的抖动。

evo_ape tum gt.txt l.txt -as  --save_plot SAVE_PLOT.pdf

  --save_plot SAVE_PLOT
                        path to save plot保存绘图的路径

$ evo_ape tum gt.txt l.txt -as --save_plot SAVE_PLOT.pdf
（略）

Plots saved to SAVE_PLOT

evo_ape tum gt.txt l.txt -as --serialize_plot SERIALIZE_PLOT

  --serialize_plot SERIALIZE_PLOT
                        path to serialize plot (experimental)序列化绘图的路径（实验）

$ evo_ape tum gt.txt l.txt -as --serialize_plot SERIALIZE_PLOT

（略）

evo_ape tum gt.txt l.txt -as --save_results ape_result.zip

  --save_results SAVE_RESULTS
                        .zip file path to store results存储结果的.zip文件路径

$ evo_ape tum gt.txt l.txt -as --save_results ape_result.zip

（略）

evo_ape tum gt.txt l.txt -as --logfile ape_result.log

  --logfile LOGFILE     Local logfile path.本地日志文件路径。

$ evo_ape tum gt.txt l.txt -as --logfile ape_result.log

（略）

可用性选项

evo_ape tum gt.txt l.txt -as  --no_warnings

 --no_warnings         no warnings requiring user confirmation无需用户确认的警告

$ evo_ape tum gt.txt l.txt -as  --no_warnings
APE w.r.t. translation part (m)
(with Sim(3) Umeyama alignment)

       max    0.384438
      mean    0.156007
    median    0.141948
       min    0.010909
      rmse    0.185972
       sse    0.795467
       std    0.101230

evo_ape tum gt.txt l.txt -v

  -v, --verbose         verbose output详细输出

$ evo_ape tum gt.txt l.txt -v
--------------------------------------------------------------------------------
Loaded 1684 stamps and poses from: gt.txt
Loaded 23 stamps and poses from: l.txt
Synchronizing trajectories...
Found 23 of max. 23 possible matching timestamps between...
    gt.txt
and:    l.txt
..with max. time diff.: 0.01 (s) and time offset: 0.0 (s).
--------------------------------------------------------------------------------
Compared 23 absolute pose pairs.
Calculating APE for translation part pose relation...
--------------------------------------------------------------------------------
APE w.r.t. translation part (m)
(not aligned)

       max    6372512.914678
      mean    6372511.473926
    median    6372511.871822
       min    6372508.722174
      rmse    6372511.473926
       sse    934004757162323.000000
       std    1.138090

-v还可以搭配-s -a -sa等一起使用，主要是输出详细数据。

evo_ape tum gt.txt l.txt -as --silent

  --silent              don't print any output不打印任何输出

$ evo_ape tum gt.txt l.txt -as --silent

 evo_ape tum gt.txt l.txt -as   --debug

  --debug               verbose output with additional debug info使用附加调试信息调试详细输出

$ evo_ape tum gt.txt l.txt -as   --debug
[DEBUG][2023-10-14 21:35:49,345][log.configure_logging():114]
System info:
Python 2.7.17
Linux-5.4.0-113-generic-x86_64-with-Ubuntu-18.04-bionic
 

[DEBUG][2023-10-14 21:35:49,345][main_ape.run():227]
main_parser config:
{'align': True,
 'align_origin': False,
 'config': None,
 'correct_scale': True,
 'debug': True,
 'est_file': 'l.txt',
 'logfile': None,
 'n_to_align': -1,
 'no_warnings': False,
 'plot': False,
 'plot_colormap_max': None,
 'plot_colormap_max_percentile': None,
 'plot_colormap_min': None,
 'plot_full_ref': False,
 'plot_mode': u'xyz',
 'pose_relation': 'trans_part',
 'ref_file': 'gt.txt',
 'ros_map_yaml': None,
 'save_plot': None,
 'save_results': None,
 'serialize_plot': None,
 'silent': False,
 'subcommand': 'tum',
 't_max_diff': 0.01,
 't_offset': 0.0,
 'verbose': False}
[DEBUG][2023-10-14 21:35:49,345][main_ape.run():228]
--------------------------------------------------------------------------------
[DEBUG][2023-10-14 21:35:49,351][file_interface.read_tum_trajectory_file():112]
Loaded 1684 stamps and poses from: gt.txt
[DEBUG][2023-10-14 21:35:49,351][file_interface.read_tum_trajectory_file():112]
Loaded 23 stamps and poses from: l.txt
[DEBUG][2023-10-14 21:35:49,351][main_ape.run():238]
Synchronizing trajectories...
[DEBUG][2023-10-14 21:35:49,352][sync.associate_trajectories():105]
Found 23 of max. 23 possible matching timestamps between...
    gt.txt
and:    l.txt
..with max. time diff.: 0.01 (s) and time offset: 0.0 (s).
[DEBUG][2023-10-14 21:35:49,353][main_ape.ape():171]
--------------------------------------------------------------------------------
[DEBUG][2023-10-14 21:35:49,353][trajectory.align_trajectory():391]
Aligning using Umeyama's method... (with scale correction)
[DEBUG][2023-10-14 21:35:49,354][trajectory.align_trajectory():403]
Rotation of alignment:
[[-0.74522462 -0.51826266 -0.41957606]
 [ 0.03226245 -0.6565123   0.75362506]
 [-0.66603258  0.5480834   0.50596955]]
Translation of alignment:
[-2853537.48321946  4667400.58781658  3268263.44928871]
[DEBUG][2023-10-14 21:35:49,354][trajectory.align_trajectory():404]
Scale correction: 0.849029484079
[DEBUG][2023-10-14 21:35:49,356][main_ape.ape():180]
--------------------------------------------------------------------------------
[DEBUG][2023-10-14 21:35:49,356][metrics.process_data():373]
Compared 23 absolute pose pairs.
[DEBUG][2023-10-14 21:35:49,356][metrics.process_data():375]
Calculating APE for translation part pose relation...
[DEBUG][2023-10-14 21:35:49,357][main_ape.ape():202]
--------------------------------------------------------------------------------
[INFO][2023-10-14 21:35:49,357][main_ape.ape():203]
APE w.r.t. translation part (m)
(with Sim(3) Umeyama alignment)

       max    0.384438
      mean    0.156007
    median    0.141948
       min    0.010909
      rmse    0.185972
       sse    0.795467
       std    0.101230