HITSZ-NRSL / Dynamic-VINS

[RA-L 2022] RGB-D Inertial Odometry for a Resource-restricted Robot in Dynamic Environments
310 stars 40 forks source link

results about evaluation #21

Open YongshuaiZ opened 1 year ago

YongshuaiZ commented 1 year ago

作者你好,我在评估算法时遇到了一些问题。我更改了visualization.cpp 中的 void pubOdometry(const Estimator &estimator, const std_msgs::Header &header)文件,更改如下,以保证它可以输出和cafe-1-1 groundtruth.txt相同类型的数据。但是我发现S使用EVO评测工具时显示 时间戳并未对齐,我想请问下这个应该如何解决。同时我可以成功使用openloris工具评测ATE,但是该工具不提供RPE的评测,想问下论文中的RPE是使用EVO评测工具吗? image

cafe-1-1-IMU.txt 评估结果 image cafe-1-1 groundtruth.txt image

wsanm commented 1 year ago

您好,您解决了吗?

YongshuaiZ commented 12 months ago

您好,您解决了吗?

还没有。同时发现使用Openloris工具对cafe1-2数据集进行评测时,跟踪准确率与论文存在差异。

wsanm commented 11 months ago

您好,您解决了吗?

还没有。同时发现使用Openloris工具对cafe1-2数据集进行评测时,跟踪准确率与论文存在差异。

您好,大概差多少呢?

YongshuaiZ commented 11 months ago

您好,您解决了吗?

还没有。同时发现使用Openloris工具对cafe1-2数据集进行评测时,跟踪准确率与论文存在差异。

您好,大概差多少呢?

可以看我另一个issue 29 主要是跟踪成功率方面,使用默认配置,成功率只有57.4%左右。

wsanm commented 11 months ago

您好,您解决了吗?

还没有。同时发现使用Openloris工具对cafe1-2数据集进行评测时,跟踪准确率与论文存在差异。

您好,大概差多少呢?

可以看我另一个issue https://github.com/HITSZ-NRSL/Dynamic-VINS/issues/29,主要是跟踪成功率方面,使用默认配置,成功率只有57.4%左右。

谢谢