You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I'm having trouble loading the TUM RGBD sequences.
I successfully converted other datasets but when converting this one the GS camera path (as seen in the viewer and training result) is wrong.
To my understanding the TUM poses are C2W poses with a COLMAP camera coordinate system (Z forward, X right and Y down). So in order to format it to a COLMAP sequence I just need to invert the poses with :
R_w2c = R_c2w.T
t_w2c = -R_w2c @ t_c2w
And get the pose in my images.txt file as quaternion and translations.
However this procedure is leading me to a wrong camera path during GS training... The camera end up under my initial gaussian representation.
It might be the same issue as in #595 but there's not much detail on the topic in this one.
Can anybody point me the thing I missed in my approach ?
Best regards,
Have a nice day,
Hugo
The text was updated successfully, but these errors were encountered:
Hi GS community,
I'm having trouble loading the TUM RGBD sequences.
I successfully converted other datasets but when converting this one the GS camera path (as seen in the viewer and training result) is wrong.
To my understanding the TUM poses are C2W poses with a COLMAP camera coordinate system (Z forward, X right and Y down). So in order to format it to a COLMAP sequence I just need to invert the poses with :
And get the pose in my
images.txt
file as quaternion and translations.However this procedure is leading me to a wrong camera path during GS training... The camera end up under my initial gaussian representation.
It might be the same issue as in #595 but there's not much detail on the topic in this one.
Can anybody point me the thing I missed in my approach ?
Best regards,
Have a nice day,
Hugo
The text was updated successfully, but these errors were encountered: