API: Identify images whose geometry was optimized together

Using the Mapillary API, is there is a way to obtain the set of images that were optimised together (their 6-DoF poses and 3D map points) within Mapillary’s Structure-from-Motion pipeline?

From the documentation, the fields sfm_cluster and merge_cc may be what I am looking for. However, they differ in what they represent. From the docs:

  • merge_cc - int, ID of the connected component of images that were aligned together.
  • sfm_cluster - { id: string, url: string } - URL to the point cloud data in JSON and compressed by zlib. See the example below

Could you please confirm if my understanding of them is correct?

  • regarding merge_cc: a local SfM reconstruction corresponds to just one merge_cc value. Thus, the 3D geometry (camera poses and 3D points) corresponding to images sharing the same merge_cc value was optimised together and no other merge_cc value exists for images participating in the same reconstruction.
  • regarding sfm_cluster: a local SfM reconstruction may contain multiple sfm_cluster values. sfm_cluster thus refers to partial local SfM reconstructions.

Any help in understanding this is really appreciated!


The following shows examples of image locations colored by merge_cc or sfm_cluster. Multiple sequences can have images with a shared merge_cc/sfm_cluster, which might indicate that a joint optimization has been done across these sequences. However, it seems that for sfm_cluster there is a stricter criterion for assigning images to the same value.

@tao do you know?