Methods assessment


(Mesheritage) #1

Hello everyone,

I am a research, even though I am not working in 3D I am using it and I am interested in method assessment to know how accurate the technique I am using is. Now my problem is I often people using a methods without proper assessment, more because "it works for them" rather than "how it should work". This is not an issue for models design to be "visual", however to optimise a process or to get accurate 3D measurment, it is important.

Most article I saw on that point are more dealing with "research algorithm" or "image analysis" (more about location on picture than proper 3D measurment).

Meanwhile in practice I encounter and heard about issue. Rounded corner through agisoft photoscan, missalignment, BA issues, wrong computation and the worst might be when we consider mesh reconstruction!

Of course one could say "if we use this or that parameter we can solve this or that issue", but the point would be to be able to know how "accuracte" we can get with a given package and given parameters as well as type of acquisition.

Hence, as there are lot of user here, I would like to collect some feedback from the community.

Did you ever do error assessment?
Did you notice some weakness/strengh of a software compare to other?

Any measurment of error versus number of picture/quality of picture?