Maxime Ferrera

What might happen is that too many images are selected to become a keyframe at some point and the queue used in the Mapper and Loop Closer gets saturated (the...

About the sleep, I indeed meant to use a usleep() function or something similar to make sure that the images are truly read at 20 Hz. About the colorization in...

Hi, Yes it is possible to add such measurements into ov2slam. First you would have to read new incoming velocity measurements in the same way that images are read in...

Yes I can provide some test data for you to validate everything. Is it OK if I email them to you (the data are not confidential but they should not...

@EwertzJN you can use the branch https://github.com/ferreram/colmap/tree/sfm_gps_ba_aligner_4_clem What you need to provide then is either images with their GPS positions set in the EXIF or update the database once you...

@EwertzJN thanks for your comments. I have been quite busy lately but I'll try to find some time to get back into this soon.

The cameras' poses in the json are the 3D transformation from world to camera. So if you want to use them to get the world coordinates of some 3D points...

Your formula is wrong, use : world_3d_pt = R_cam2world * cam_3d_pt + t_cam2world

In point_triangulator, neither intrinsics, distortion nor extrinsics are optimized. The fact that they are all kept fixed is hard-coded (check RunPointTriangulatorImpl function). The point_triangulator step is performing triangulation of matched...

Hi, I think your problem is due to the principal point being given as an int instead of a double. Try this : colmap feature_extractor --database_path db.db --image_path images --ImageReader.camera_model...