I've been using the following set of commands to result in a dense reconstruction on a docker volume:
- colmap automatic_reconstructor --image_path "$DOCKER_VOLUME/raw_frames/" --workspace_path $DOCKER_VOLUME/colmap --single_camera 1
- colmap image_undistorter --image_path $DOCKER_VOLUME/raw_frames --input_path $DOCKER_VOLUME/colmap/sparse/0 --output_path $DOCKER_VOLUME/colmap/dense --output_type COLMAP
- colmap patch_match_stereo --workspace_path $DOCKER_VOLUME/colmap/dense --workspace_format COLMAP --PatchMatchStereo.geom_consistency true
Unfortunately, I am getting very different, poorer results from the command line interface than when using the colmap gui automatic reconstruction application. The only non-default option I've selected in the gui is to select "shared intrinsics", which I've tried to do via command line with "--single-camera 1", but this doesn't seem to be working. Even my sparse reconstruction from the command line is a lot worse.
Any suggestions on how to improve the quality of my reconstruction from the command line?