search

mapillary / OpenSfM

Open source Structure-from-Motion pipeline
https://www.opensfm.org/
BSD 2-Clause "Simplified" License
3.35k stars 852 forks source link

Creating Submodels #687

Open 2016ee165 opened 3 years ago

2016ee165 commented 3 years ago

is it possible to create submodels without having the GPS position of the images?

Talha-Hammad commented 3 years ago

if you would like to manually do it, you can pass your own list of image neighbor pairs in line 45 before returning https://github.com/mapillary/OpenSfM/blob/d39c1c1e1dbf2cfe4a704fc1293812a5b5d75c94/opensfm/matching.py#L38-L50 the rest of the pipeline will work as expected. This will not create subdirectories for submodels but will limit your matching and potentially save you a lot of time.

2016ee165 commented 3 years ago

@Talha-Hammad my question wasn't about matching image pairs. I want to split my dataset into smaller datasets, run opensfm on all the datasets separately and then combine them at the end. In the documentation, there is a method to achieve this task but it uses GPS information of images. I wanted to do so without the GPS information of images.

CodeeCat commented 3 years ago

There is no feedback from the submodel command after the data set with GPS information is rebuilt . root@ubuntu:~/openSfm/OpenSfM# bin/opensfm align_submodels /data/shitou 2021-02-07 16:10:30,643 INFO: Running alignment 2021-02-07 16:10:30,643 INFO: Ceres Solver Report: Iterations: -2, Initial cost: 0.000000e+00, Final cost: 0.000000e+00, Termination: CONVERGENCE Do you have this problem？

benjaminangerer commented 2 years ago

Is there any progress or workaround for this issue? We have the same error message here and we really would like to use the submodels feature.