I found some issues with
bob bio gen.
First, the probes are inconsistently assigned to wrong subject ids. For instance, if you do
bob bio gen scores you will have lines like this
x y s5 8.193197 and
x x s5 10.347785 where the probe
s5 belongs to the
y at the same time.
bob bio gen is limited to two subjects only. This is ok when you want to plot a ROC, DET, EPC or histograms, but is not ok if you want to generate a CMC plot.
Finally, it is not possible to generate scores for open set scenarios, where there are probes from unknown clients (with no associated model).
This MR solves all these aforementioned issues.
If you do
bob bio gen --help you will see:
Usage: bob bio gen [OPTIONS] OUTDIR Generate random scores. Generates random scores in 4col or 5col format. The scores are generated using Gaussian distribution whose mean is an input parameter. The generated scores can be used as hypothetical datasets. Options: -mm, --mean-match FLOAT Mean for the positive scores distribution [default: 10] -mnm, --mean-non-match FLOAT Mean for the negative scores distribution [default: -10] -p, --n-probes-per-subjects INTEGER Number of probes per subject [default: 5] -s, --n-subjects INTEGER Number of subjects [default: 5] -p, --sigma-positive FLOAT Variance for the positive score distributions [default: 10] -n, --sigma-negative FLOAT Variance for the negative score distributions [default: 10] -u, --n-unknown-subjects INTEGER Number of unknown subjects (useful for openset plots) [default: 0] --five-col / --four-col [default: False] -v, --verbose Increase the verbosity level from 0 (only error messages) to 1 (warnings), 2 (log messages), 3 (debug information) by adding the --verbose option as often as desired (e.g. '-vvv' for debug). -?, -h, --help Show this message and exit.