The paper says that "The framework uses the default values of the configuration parameters of any underlying engines such as feature extraction, disparity calculations", but the disparity parameters (i.e. for SGBM) seem to be manually set and different than the opencv default:
-32, // MinDisparity
144, // NumDisparities
9, // block size
60, // P1
2400, // P2
90, // Disp12MaxDiff
16, // PreFilterCap
1, // UniquenessRatio
60, // SpeckleWindowSize
20, // SpeckleRange
true // SGBM mode
are these from an example somewhere, and if not, how did you arrive at these parameters?

Additionally, the disparity output seems to be normalized w/ NORM_MINMAX before the statistics are collected. Unless the min/max pixel values are always 0,255 will this not degrade the value of comparing the disparity mean and stddev across frames within an input or across input sequences?

Hello,
Thanks for the great work!
Ran it out of the box using the docker instructions and this is kind of how that went:

Unless the report filepath exists, the analysis data isn't written to the reports. Something like adding the line in kitti_min.sh to the other scripts fixed it and allowed the reports to be generated: mkdir ../out_data/reports/KITTI

Characterization option failure:
[g] (general) just immediately segfaults for me
[b] (inertial diff) segfaults
^I don't know what to attribute this to except some issue in the docker build...I remember seeing some red text in the docker build but don't remember what it was

[3] (brightness) gets to 99% for the first input then segfaults...maybe some aggregation or file writing error?
[a] (inertial dynamic range) segfaults with a floating point exception...maybe a divide by 0? Not sure.
^Will probably try and meander my way through these errors in the next couple of weeks.

Thought I'd bring up the part about the reports ASAP and the rest is in case it's useful somehow.

My system is ubuntu 20.04 and docker version is: Docker version 20.10.21, build baeda1f. The dataset I've been working with is euroc