Pos.txt parameters ??

[makaay2077]

Hi everyone

would like to ask you something about this project

How do we create pos.txt data in the dataset you use in the SphereSfm project? What parameters are in it? I also have trajectory data, but I don't know which parameters to use and how. Can you help me?

[json87]

pos.txt contains the ground-truth data of image poses by using IMU.

[makaay2077]

Thanks for reply

Do you know what the columns in the data are? In my own dataset I have a position file with columns "Name,Latitude,Longitude,AbsAlt,RltAlt,Pitch,Roll,Yaw,StdHorizon,StdHeight,Q,Date,Time". Can I convert this to the format that will be used in colmap?

[json87]

"Name,Latitude,Longitude,AbsAlt,Pitch,Roll,Yaw“ is the data you should use. Besides, you should convert "Pitch,Roll,Yaw" in the navigation system to the computer vision Or photogrammeric system.

[yt2639]

Hii @json87 , thank you for providing the code.

I understand (n,x,y,z,R) = "Name,Latitude,Longitude,AbsAlt,Pitch,Roll,Yaw“ is what we need for the POS.txt. However, in my case, I don't seem to have this GT-style info. What I have is just world-to-camera transformation matrix for each ERP image (namely I already know the poses). Is it appropriate to use w2c[3,:3] as x,y,z and w2c[:3,:3] as R?

And generally, in this case, how can I incorporate the known poses into this pipeline?

Thank you!

[json87]

Hii @json87 , thank you for providing the code.

I understand (n,x,y,z,R) = "Name,Latitude,Longitude,AbsAlt,Pitch,Roll,Yaw“ is what we need for the POS.txt. However, in my case, I don't seem to have this GT-style info. What I have is just world-to-camera transformation matrix for each ERP image (namely I already know the poses). Is it appropriate to use w2c[3,:3] as x,y,z and w2c[:3,:3] as R?

And generally, in this case, how can I incorporate the known poses into this pipeline?

Thank you!

w2c[:3,:3] is ok for R w2c[:3,3] is the translation vector T, which should be converted with C = -R.transpose()*T.