Obvious features doesn't match

[Mete-cell]

I can't get get the obvious matches. Dataset contains 141 equirectangular pictures á 16384x8192 with exif gps coordinates. I'm using the SPHERE Camera Model.

Generated a pos_path textfile with the average latitude and longitude as the origin before transforming the gps coordinates to enu.

#F=N X Y Z
image001.jpg 95.606 17.457 -0.001
image002.jpg 101.553 14.676 -0.001
image003.jpg 110.377 10.349 -0.001
image004.jpg 122.078 4.477 -0.001
image005.jpg 130.902 0.151 -0.001
image006.jpg 139.726 -4.176 -0.002
image007.jpg 149.894 -27.664 -0.002
image008.jpg 157.374 -12.829 -0.002
image009.jpg 169.076 -18.701 -0.002
image010.jpg 175.023 -21.482 -0.002
image011.jpg 183.847 -25.809 -0.003
image012.jpg 192.671 -30.135 -0.003
...

Set max_num_inliers to 50.

Does anyone have an idea?

[ahojnnes]

The normal version of colmap does not have a spherical camera model implemented. Which version are you using?

[Mete-cell]

The normal version of colmap does not have a spherical camera model implemented. Which version are you using?

I'm using the binaries of SphereSfM. https://github.com/json87/SphereSfM

[sarlinpe]

We can't provide support for forks of COLMAP so I suggest opening this issue there. FYI I expect that COLMAP will soon support such spherical images.