I've been evaluating the behavior of GLOMAP when modifying some of the input parameters. I was surprised to find that, for some parameters, the performance of GLOMAP doesn't change at all, even when I input highly degraded configurations. For example:
After digging into the code, I suspect that some of these parameters are not being properly parsed. For instance, in glomap/glomap/estimators/bundle_adjustment.h:
The thres_loss_function value is hard-coded there. Additionally, I suspect that even if I comment out that line, it won't work since the loss_function is set up before the thres_loss_function is parsed in exe/global_mapper.cc.
Could you help me confirm if my understanding is correct?
Hi,
I've been evaluating the behavior of GLOMAP when modifying some of the input parameters. I was surprised to find that, for some parameters, the performance of GLOMAP doesn't change at all, even when I input highly degraded configurations. For example:
After digging into the code, I suspect that some of these parameters are not being properly parsed. For instance, in
glomap/glomap/estimators/bundle_adjustment.h
:The
thres_loss_function
value is hard-coded there. Additionally, I suspect that even if I comment out that line, it won't work since theloss_function
is set up before thethres_loss_function
is parsed inexe/global_mapper.cc
.Could you help me confirm if my understanding is correct?
Thanks,