Analysis of the results from reconstruction pipeline

As someone new to the OpenSfM project, I am still not sure how to analyze the results produced from the reconstruction pipeline. I see different reports are being generated when I run something like opensfm_run_all

But I would love to know, What features are good and have contributed to the reconstruction and what are bad features, developing a qualitative basis to filter the good and bad features. At the same time, filtering out descriptors is the same way to quantify something like - ‘good feature descriptor’ and ‘good features’.

Any leads or directions to analyze results quantitatively and qualitatively will be helpful. Later, potentially validating against the ground truth.

End goal: To create a good Feature descriptor and feature library and later save it in some database like - SQLite.

Thanks in advance for your help.


This post is best suited to the Data subsection of the forum