Open pierotofy opened 3 years ago
I'd lean toward aggregate stats. Any reason to not have submodel stats as well? For larger datasets it could be more useful than global stats.
I agree to both. The bottom line is that the split-merge report would contain slightly different information compared to normal runs.
Yup. Agreed.
@pierotofy opensfm/action/compute_statistics.py
calls stats.py
functions that works on a list of reconstructions.
Didn't know that. Thanks @YanNoun ! I'll try it out.
The save_heatmap
/save_topview
might be super slow with 100K/1M points, feel free to run some random index-based decimation.
I did notice some slowdowns with larger datasets; would this be a good idea for a PR to OpenSfM for the compute_statistics
command? (I could add a --decimate
flag)
Yup definitely, that would be great !
https://github.com/mapillary/OpenSfM/pull/695
I opted to allow a user to set a cap on the number of points using random sampling. Definitely helps keep runtime in check!
Currently ODM does not generate PDF reports when using split-merge. It could be possible to generate them by: