Hi guys,
I have scenario for EMG acquisition and classification that's been working properly for months now. However in the last week or so, OpenVibe keeps reporting next errors:
[ ERROR ] At time 131.000 sec <Box algorithm::(0x000072db, 0x00005828) aka Classifier trainer> stopIndex-trainIndex=1
[ ERROR ] At time 131.000 sec <Box algorithm::(0x000072db, 0x00005828) aka Classifier trainer> Bailing out (from xval)...
[WARNING] Box algorithm <Classifier trainer> has been deactivated because process() function returned error
I haven't changed anything in the programs, and have tried different versions, all with the same error.
Could you be able to help me, please?
Box Error
Re: Box Error
Hi,
I need some more informations to be able to help you.
Which version of OpenViBE do you use?
What is the k value for the k-fold?
Cheers,
Guillaume
I need some more informations to be able to help you.
Which version of OpenViBE do you use?
What is the k value for the k-fold?
Cheers,
Guillaume
Re: Box Error
Hi,
I have the same error when trying to use a classifier trainer.
[ ERROR ] At time 62.844 sec <Box algorithm::(0x00004eb7, 0x00000d37) aka Classifier trainer> stopIndex-trainIndex=1
[ ERROR ] At time 62.844 sec <Box algorithm::(0x00004eb7, 0x00000d37) aka Classifier trainer> Bailing out (from xval)...
[WARNING] Box algorithm <Classifier trainer> has been deactivated because process() function returned error
I have OpenVibe 1.0.0 and i have other scenarios working perfectly well with classifier trainers. I tried a lot of different k values to try to fix my problem but it didn't work. I use the exact same settings for a classifier trainer in an other scenario (except save path) and i encounter no problem so i really don't get why it doesn't work in that case. I have nearly the same structure as the classification training given as example in the "openvibe\share\openvibe\scenarios" folder so everything looks alright to me.
Any idea what could cause this error ?
I will also try with other stream and stimulations but i don't think it's the origin of the problem.
Thanks for help,
Mathilde
I have the same error when trying to use a classifier trainer.
[ ERROR ] At time 62.844 sec <Box algorithm::(0x00004eb7, 0x00000d37) aka Classifier trainer> stopIndex-trainIndex=1
[ ERROR ] At time 62.844 sec <Box algorithm::(0x00004eb7, 0x00000d37) aka Classifier trainer> Bailing out (from xval)...
[WARNING] Box algorithm <Classifier trainer> has been deactivated because process() function returned error
I have OpenVibe 1.0.0 and i have other scenarios working perfectly well with classifier trainers. I tried a lot of different k values to try to fix my problem but it didn't work. I use the exact same settings for a classifier trainer in an other scenario (except save path) and i encounter no problem so i really don't get why it doesn't work in that case. I have nearly the same structure as the classification training given as example in the "openvibe\share\openvibe\scenarios" folder so everything looks alright to me.
Any idea what could cause this error ?
I will also try with other stream and stimulations but i don't think it's the origin of the problem.
Thanks for help,
Mathilde
Re: Box Error
Hi Mathilde,
I will need more informations in order to help you. Can you upload the scenario which give you that error?
Guillaume
I will need more informations in order to help you. Can you upload the scenario which give you that error?
Guillaume
Re: Box Error
Hello,
Sorry for the late reply... I'm using latest version of openvibe (20/03/15) as I updated the previous one (and have updated all the boxes).
The k-fold value is 7.
Please find attached scenarios that I've been using for the handball demo.
The correct order of application is handball emg acquisition, then classifier trainer and at the end: online.
Thanks a lot
Tijana
Sorry for the late reply... I'm using latest version of openvibe (20/03/15) as I updated the previous one (and have updated all the boxes).
The k-fold value is 7.
Please find attached scenarios that I've been using for the handball demo.
The correct order of application is handball emg acquisition, then classifier trainer and at the end: online.
Thanks a lot
Tijana
- Attachments
-
- handball-emg-online.xml
- (47.85 KiB) Downloaded 284 times
-
- handball-emg-classifier-trainer.xml
- (51.68 KiB) Downloaded 265 times
-
- handball-emg-acquisition.xml
- (29.98 KiB) Downloaded 277 times
Re: Box Error
Hi,
Could you please put the log you get during execution of the scenario. (Paste them in a file and upload it)
Guillaume
Could you please put the log you get during execution of the scenario. (Paste them in a file and upload it)
Guillaume
Re: Box Error
Hello Guillaume,
Please find a copy/paste in a .txt file of the errors provided while running the scenarios previously attached.
Please let me know if there's anything else I can provide for you.
Cheers,
Tijana
Please find a copy/paste in a .txt file of the errors provided while running the scenarios previously attached.
Please let me know if there's anything else I can provide for you.
Cheers,
Tijana
- Attachments
-
- OpenVibeLog.txt
- (1.73 KiB) Downloaded 279 times
Re: Box Error
Hi,
According to the log, the trainer box only received 10 feature vectors. Is that normal?
Guillaume
According to the log, the trainer box only received 10 feature vectors. Is that normal?
Guillaume
Re: Box Error
Hello,
Thanks for the suggestion, now there's no error.
However, when the classification process is done, the third scenario only reacts to one class, not the other one.
For example, if one class is resting and the other EMG, it only recognize resting and not EMG.
What could be the problem?
Thanks,
Tijana
Thanks for the suggestion, now there's no error.
However, when the classification process is done, the third scenario only reacts to one class, not the other one.
For example, if one class is resting and the other EMG, it only recognize resting and not EMG.
What could be the problem?
Thanks,
Tijana
Re: Box Error
Hi,
For me the amount of data is a problem.
Could you try to have more samples of each class?
Guillaume
For me the amount of data is a problem.
Could you try to have more samples of each class?
Guillaume