Closed zhangChen341 closed 5 months ago
You can refer to the paper listed in the repo page "3D reconstruction of spherical images based on incremental structure from motion". The ground-truth data is captured by using on-board high-precision POS system of the MMS.
My dataset is took from Insta360 X2 camera ,
How can i have the rotation data , Will there a seperate system to capture spatial rotations of each frame , I have x,y,z with me but i am not able to get rotations to keep in Pos.txt
Can you help me with that ?
or
Any guide to get them and use them in our dataset ! As the spatial gives good results compared to Exhaustive & Vocab tree
My dataset is took from Insta360 X2 camera ,
How can i have the rotation data , Will there a seperate system to capture spatial rotations of each frame , I have x,y,z with me but i am not able to get rotations to keep in Pos.txt
Can you help me with that ?
or
Any guide to get them and use them in our dataset ! As the spatial gives good results compared to Exhaustive & Vocab tree
For feature matching, XYZ is enough as the angles are not used in the spatial matching algorithm. Thus, you can define R as an identity matrix in the POS.txt
Okay ! But still I am loosing some poses !
Check this !
Can i know reasons where it may fail ! Like that , I thought Rotations may be one reason for that ! Is that ??
Hello,
I'm working with coordinate data (POS.txt ) that includes spatial positions and a rotation array, as outlined in the following format: #F=N X Y Z R An example entry from the data is as follows:
Example For an image in POS.txt :
Job_202306201252_Track01_Sphere_00001.jpg 182.0801976 536.9640277 5.533966177 0.979997362 -0.011776529 0.198661731 0.19834921 -0.023510662 -0.979849396 0.016209893 0.999654221 -0.020704522
How can I get :
R = [0.979997362 -0.011776529 0.198661731 0.19834921 -0.023510662 -0.979849396 0.016209893 0.999654221 -0.020704522 ]
Please Can i Know,
Thank you for your assistance!
Best regards, Zhang