search

raulmur / ORB_SLAM2

Real-Time SLAM for Monocular, Stereo and RGB-D Cameras, with Loop Detection and Relocalization Capabilities
Other
9.3k stars 4.7k forks source link

Some questions about system performance #883

Open SakuraMemoryKnight opened 4 years ago

SakuraMemoryKnight commented 4 years ago

I read some of the papers on slam, and I found that in these papers the performance of slam was usually evaluated with the average relative translation （%） and rotation （deg/m）, It's just like the kitti list shows： 77239946-483b8b00-6c1b-11ea-8c01-f2276333e6ae How to get this result? In fact, I don't quite understand the meaning of these two evaluation indexes, so I don't know what the formula is for this parameter. I have used evo to test the slam performance results, and can get ape and rpe, but these are not the above two evaluation parameters. I just want to know what their formula is. Is it the absolute error of the final frame divided by the total distance？

bhaveshachhada commented 4 years ago

@raulmur , If I have generated my own dataset by moving a RGBD camera in my backyard of in my house, If I run ORB SLAM 2, How can I measure the accuracy ?