bob.spear 1.1.8 and 1.1.7

54 views
Skip to first unread message

Maria Garcia

unread,
Mar 12, 2015, 5:33:19 PM3/12/15
to bob-...@googlegroups.com
Hi,
What are the improvements/changes between 1.1.7 and 1.1.8?

I am considering upgrading to 1.1.8 but cannot seem to find a list of the changes anywhere.  a friend tried to upgrade but had trouble and I'm hesitant to do so unless there are some speedups or big changes to accuracy.

Thank you

Maria

Manuel Günther

unread,
Mar 13, 2015, 5:59:31 AM3/13/15
to bob-...@googlegroups.com
Dear Maria,

the difference between 1.1.7 and 1.1.8 is purely fixing some installation issues. Since some of the packages (like facereclib in version > 2.0) is Bob 2 compatible only, we needed to pin-point versions of these packages that work together with Bob 1.

On the other hand, there is no speed-ups or accuracy changes. So, as long as 1.1.7 is working well in your environment, there is no need to upgrade.

Best regards
Manuel

Maria Garcia

unread,
Mar 13, 2015, 11:41:50 AM3/13/15
to bob-...@googlegroups.com
Hi Manuel,
Thanks for the information.   Do you have any documentation or backgrounders on what I can do to make my results more accurate?   For example I am running:
./bin/spkverif_ivector.py -d config/database/voxforge.py -p config/preprocessing/energy.py \
     -f config/features/mfcc_60.py   -t config/tools/ivec/ivec_256g_t100_cosine.py -z -b ivector_cosine \
     --groups eval   --skip-lda-train-projector --skip-lda-projection \
     --skip-wccn-train-projector --skip-wccn-projection --skip-train-plda-enroler \
     --skip-model-enrolment

I'm using my own data with the voxforge protocol but the rest of the settings are default.  I don't really know very much about speaker ID and am learning this as I go.   I'd like to know what settings I can change to perhaps make things more accurate.   For example if I were to add more data to my UBM, would that  help?  right now I have about 500 speakers in the UBM with  8-9 seconds of speech per speaker.   Would I have better accuracy if I had 2000 speakers at 8-9 seconds?  or if I had 500 speakers at 30 seconds?  None of the speakers in the UBM are part of the enroll or tests.

Thanks,
Maria

--
-- You received this message because you are subscribed to the Google Groups bob-devel group. To post to this group, send email to bob-...@googlegroups.com. To unsubscribe from this group, send email to bob-devel+...@googlegroups.com. For more options, visit this group at https://groups.google.com/d/forum/bob-devel or directly the project website at http://idiap.github.com/bob/
---
You received this message because you are subscribed to the Google Groups "bob-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email to bob-devel+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Reply all
Reply to author
Forward
0 new messages