Issue when generating features from APT

[2909ft]

Hello everybody!

I am trying to create a new JAABA project using the tracks from APT but I am running into some issues when generating features. (see screenshot) The file is present inside the experiment directory, I don't understand why JAABA cannot find it. Any ideas on how to solve this?

Also, I have another question: is it possible to create an experiment using only the tracks from APT? At the moment, I still track the video using motr, and then create an experiment based on that using PrepareJAABAData, and only then I transfer the trk file from APT to the experiment directory. Is there a more efficient way of doing this? Am I doing it completely wrong?

Kind regards,
Filipa

[mkabra]

Hi Filipa, I cannot see the full name of the trk file in the screenshot. Can you ensure that the full file name of the trk file is same as the one reported in the error message? The trk file on disk probably has the tracker name ("mdn" or "dlc") at the end. When you create the project in JAABA, you should give this name. JAABA does pre-fill the trk field, but it isn't always entirely correct. Your workflow is mostly correct for projects with multiple animals. The way I would do it is: 1) Track using MoTR. 2) Convert MoTR output to trx. However, you can avoid using PrepareJAABAData GUI for this. Check here: https://groups.google.com/forum/#!searchin/jaaba/preparejaabadata%7Csort:date/jaaba/ecuN2hBr-AY/ukJJo2q8CQAJ 3) Do APT tracking on the movie and trx file in the above generated JAABA compatible directory. That way the trk file will get generated in the same directory. This hopefully should reduce manual intervention. Mayank

…

On Thu, Feb 13, 2020 at 4:15 PM 2909ft ***@***.***> wrote: Hello everybody! I am trying to create a new JAABA project using the tracks from APT but I am running into some issues when generating features. (see screenshot) The file is present inside the experiment directory, I don't understand why JAABA cannot find it. Any ideas on how to solve this? Also, I have another question: is it possible to create an experiment using only the tracks from APT? At the moment, I still track the video using motr, and then create an experiment based on that using PrepareJAABAData, and only then I transfer the trk file from APT to the experiment directory. Is there a more efficient way of doing this? Am I doing it completely wrong? Kind regards, Filipa [image: image] <https://urldefense.com/v3/__https://user-images.githubusercontent.com/57539512/74425614-47614d80-4e54-11ea-834e-b2cf80a52eee.png__;!!Eh6p8Q!WiI0bUy94erqVM8B190DQC0IiGYvvKNF59qzBQ_hj9uoR71VSbEKjfvtWeIOvkDvNw0$> — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <https://urldefense.com/v3/__https://github.com/kristinbranson/JAABA/issues/329?email_source=notifications&email_token=AAJNKYZK4NFE3NM2Y4IGXSLRCUQEZA5CNFSM4KUPS6EKYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4INHA4HA__;!!Eh6p8Q!WiI0bUy94erqVM8B190DQC0IiGYvvKNF59qzBQ_hj9uoR71VSbEKjfvtWeIOXbTIXww$>, or unsubscribe <https://urldefense.com/v3/__https://github.com/notifications/unsubscribe-auth/AAJNKY76LRWVBRUAW52RVOTRCUQEZANCNFSM4KUPS6EA__;!!Eh6p8Q!WiI0bUy94erqVM8B190DQC0IiGYvvKNF59qzBQ_hj9uoR71VSbEKjfvtWeIO2eiU6Ss$> .

[2909ft]

Dr. Mayank,

Thank you! Indeed, there was a sneaky dlc at the end of the file name, I completely missed it. It is wokrking fine now, and thanks for the tips for increasing efficiency.

Filipa

[2909ft]

Also, another basic question related to APT: I cannot find the option to track all movies at once, only the current movie, the dropdown menu does not give me that option. Do I have to switch movies everytime I want to track a different one?

[allenleetc]

Hi Filipa,
Within the UI you are correct, you need to switch movies at the moment to track a new movie. The pulldown menu used to include options to track "all movies" but the menu got busy and hard to manage. How many movies do you need to track, and which tracker/backend are you using? If it is a large number of movies, we do have some "Production Tracking" facilities available for bulk tracking. Currently these are available at the command-line and not yet in the UI.
Allen

[2909ft]

Dear Allen,

Currently I am just testing it, so only 5, but in the future I should have around 50 new videos every 2 weeks. How would I access these "production tracking" facilities?

Kind regards,
Filipa

[allenleetc]

Hi Filipa,

We will be documenting this soon and are also deliberating on some related updates. Do you need this immediately? If not please see kristinbranson/APT#263 where we will post updates. We should be getting to this fairly soon.

Thanks, Allen

[2909ft]

Hey Allen! Thanks for the reply!

How soon? Beginning of March?

Thanks, Filipa

[arobie]

Hi Filipa I wrote up some directions for bulk tracking: https://github.com/kristinbranson/APT/wiki/Tracking-Movies-(DL-algorithms) Sounds like you'll probably want to use option #2, but I think we'll also be adding a 'track all movies' option back into the GUI. Please let me know if they aren't clear! Cheers Alice

…

On Mon, Feb 17, 2020 at 3:25 AM 2909ft ***@***.***> wrote: Hey Allen! Thanks for the reply! How soon? Beginning of March? Thanks, Filipa — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#329?email_source=notifications&email_token=AAISKEAUEGQBJPFTPIW7273RDJCWZA5CNFSM4KUPS6EKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEL5PFVA#issuecomment-586871508>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAISKECZHN3SAZSVEONI42TRDJCWZANCNFSM4KUPS6EA> .

[2909ft]

Dear Alice!

Thank you very much! The instructions are super clear!

Cheers,
Filipa

[2909ft]

Hello! Sorry to bother again!

I tried running it, but I get the following errors:

Index exceeds the number of array elements (1).

Error in DLBackEndClass/getFreeGPUs (line 288)
freemem = freemem(1:nrequest);

Error in DeepTracker/getFreeGPUs (line 532)
[gpuid,freemem,gpuInfo] = be.getFreeGPUs(nrequest,varargin{:});

Error in DeepTracker/track (line 1949)
gpuids = obj.getFreeGPUs(nmovies*nviews);

Any idea on how to fix it? I trying to run this:

lObj.tracker.track({'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_2.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_3.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_5.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_6.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_11.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_14.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_17.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_18.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_34.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_67.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_73.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_75.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_77.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_80.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_85.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_87.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_92.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_94.avi'},'trkfiles',{'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_2.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_3.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_5.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_6.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_11.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_14.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_17.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_18.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_34.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_67.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_73.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_75.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_77.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_80.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_85.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_87.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_92.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_94.trk'})

[arobie]

This looks like an APT issue, with tracking? I'll pass it along. Alice

…

On Wed, Feb 19, 2020 at 8:56 AM 2909ft ***@***.***> wrote: Hello! Sorry to bother again! I tried running it, but I get the following errors: Index exceeds the number of array elements (1). Error in DLBackEndClass/getFreeGPUs (line 288) freemem = freemem(1:nrequest); Error in DeepTracker/getFreeGPUs (line 532) [gpuid,freemem,gpuInfo] = be.getFreeGPUs(nrequest,varargin{:}); Error in DeepTracker/track (line 1949) gpuids = obj.getFreeGPUs(nmovies*nviews); Any idea on how to fix it? I trying to run this: lObj.tracker.track({'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_2.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_3.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_5.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_6.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_11.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_14.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_17.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_18.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_34.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_67.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_73.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_75.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_77.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_80.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_85.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_87.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_92.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_94.avi'},'trkfiles',{'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_2.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_3.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_5.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_6.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_11.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_14.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_17.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_18.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_34.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_67.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_73.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_75.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_77.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_80.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_85.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_87.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_92.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_94.trk'}) — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#329?email_source=notifications&email_token=AAISKEEVDFOEH3UYAMDJQFTRDU3BNA5CNFSM4KUPS6EKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEMH6G7Q#issuecomment-588243838>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAISKEFMXMKKU66SB3GT5YTRDU3BNANCNFSM4KUPS6EA> .

[arobie]

What GPU backend are you using? Alice

…

On Wed, Feb 19, 2020 at 9:07 AM Alice Robie ***@***.***> wrote: This looks like an APT issue, with tracking? I'll pass it along. Alice On Wed, Feb 19, 2020 at 8:56 AM 2909ft ***@***.***> wrote: > Hello! Sorry to bother again! > > I tried running it, but I get the following errors: > > Index exceeds the number of array elements (1). > > Error in DLBackEndClass/getFreeGPUs (line 288) > freemem = freemem(1:nrequest); > > Error in DeepTracker/getFreeGPUs (line 532) > [gpuid,freemem,gpuInfo] = be.getFreeGPUs(nrequest,varargin{:}); > > Error in DeepTracker/track (line 1949) > gpuids = obj.getFreeGPUs(nmovies*nviews); > > Any idea on how to fix it? I trying to run this: > > > lObj.tracker.track({'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_2.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_3.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_5.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_6.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_11.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_14.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_17.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_18.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_34.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_67.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_73.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_75.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_77.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_80.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_85.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_87.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_92.avi';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_94.avi'},'trkfiles',{'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_2.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_3.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_5.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_6.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_11.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_14.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_17.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_18.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_34.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_67.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_73.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_75.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_77.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_80.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_85.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_87.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_92.trk';'Z:\Filipa\Bonsai_Chr2_Videos_Separated_Trials\m2sensoryD2_94.trk'}) > > — > You are receiving this because you commented. > Reply to this email directly, view it on GitHub > <#329?email_source=notifications&email_token=AAISKEEVDFOEH3UYAMDJQFTRDU3BNA5CNFSM4KUPS6EKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEMH6G7Q#issuecomment-588243838>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AAISKEFMXMKKU66SB3GT5YTRDU3BNANCNFSM4KUPS6EA> > . >

[2909ft]

Dear Alice,

I am using conda

Filipa

[2909ft]

It passes the configuration test, if that helps

[2909ft]

Also, another issue (this one more related to JAABA itself): I am trying to add new experiments to a jaaba classifier, but when I do so, JAABA tries to use the original trk file (the one I initally used when creating the classifier) to generate the new features on the perframe directory. However, the new experiments have different trk files, which means that JAABA is not be able to find the original trk, giving me an error message. Is there a way to get around this?

[arobie]

Not sure I total understand. Do you have movies with trk files from different trackers or just different trk file names from the same tracker? If its the first, I wouldn't recommend this because it might cause a systematic difference in the perframe features for the different movies. You may need to delete the previously calculated perframe features if you change the trackersto force recalculation of the features. I think we should add a check for this. If its just a naming difference, you can either make a new jab file that points to the correct trk file name (you can import movies and labels from old jab file). Or you can rename the trk files to match what JAABA expects. JAABA is not setup to take trk files with different names for different movies. Alice

…

On Wed, Feb 19, 2020 at 11:40 AM 2909ft ***@***.***> wrote: Also, another issue (this one more related to JAABA itself): I am trying to add new experiments to a jaaba classifier, but when I do so, JAABA tries to use the original trk file (the one I initally used when creating the classifier) to generate the new features on the perframe directory. However, the new experiments have different trk files, which means that JAABA is not be able to find the original trk, giving me an error message. Is there a way to get around this? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#329?email_source=notifications&email_token=AAISKEHM2LNPGPECZJYMJSTRDVOHVA5CNFSM4KUPS6EKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEMIQBBQ#issuecomment-588316806>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAISKEGND4NDWZLKAG5VJL3RDVOHVANCNFSM4KUPS6EA> .

[2909ft]

Dear Alice, Thank you for the reply! :) When I export the tracking results for my 6 videos, I get 6 different trk files (with different names, that is), but all from the same tracker type. What I was doing until now was editing the jaaba project before opening on training mode and importing the features from APT again, and repeating this process for each video that I wanted to add to that project. I was just wondering if there was a faster way of doing this, since this will take quite some time to do when the amount of videos I have increases substantially. Kind regards, Filipa A quarta, 19/02/2020, 18:44, arobie <[email protected]> escreveu:

…

Not sure I total understand. Do you have movies with trk files from different trackers or just different trk file names from the same tracker? If its the first, I wouldn't recommend this because it might cause a systematic difference in the perframe features for the different movies. You may need to delete the previously calculated perframe features if you change the trackersto force recalculation of the features. I think we should add a check for this. If its just a naming difference, you can either make a new jab file that points to the correct trk file name (you can import movies and labels from old jab file). Or you can rename the trk files to match what JAABA expects. JAABA is not setup to take trk files with different names for different movies. Alice On Wed, Feb 19, 2020 at 11:40 AM 2909ft ***@***.***> wrote: > Also, another issue (this one more related to JAABA itself): I am trying > to add new experiments to a jaaba classifier, but when I do so, JAABA tries > to use the original trk file (the one I initally used when creating the > classifier) to generate the new features on the perframe directory. > However, the new experiments have different trk files, which means that > JAABA is not be able to find the original trk, giving me an error message. > Is there a way to get around this? > > — > You are receiving this because you commented. > Reply to this email directly, view it on GitHub > < #329?email_source=notifications&email_token=AAISKEHM2LNPGPECZJYMJSTRDVOHVA5CNFSM4KUPS6EKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEMIQBBQ#issuecomment-588316806 >, > or unsubscribe > < https://github.com/notifications/unsubscribe-auth/AAISKEGND4NDWZLKAG5VJL3RDVOHVANCNFSM4KUPS6EA > > . > — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#329?email_source=notifications&email_token=ANW7XOCF5T6VEVDFZVH2YDTRDVVWVA5CNFSM4KUPS6EKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEMIYHNA#issuecomment-588350388>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ANW7XOCMN7ME5BX4NSRMGYTRDVVWVANCNFSM4KUPS6EA> .

[allenleetc]

Hey Filipa,

I reproduced your issue with tracking. We can fix this but it will take a little time. I have noted it at kristinbranson/APT#263.

The production tracking stuff with the local backends (Conda/Docker) is newer and has not received as much testing. Alice had noted this as well we just hadn't gotten around to it.

If you are able to track for now in the UI, we can shoot to have the fix in March if that works. Let us know, thanks.

[arobie]

JAABA expects data to be organized at the directory level, with each experiment directory (unique name for experiment) containing files with the same name. For example: /fuzzybunny_20200201/trx.mat /fuzzybunny_20200201/trackingfile.trk /fuzzybunny_20200201/movie.avi /fuzzybunny_20200201/perframe/velmag.mat see http://jaaba.sourceforge.net/DataFormatting.html#ExperimentDirectoryStructure Since you already have to reorganize your data into this format for JAABA, I would do so before tracking and then just output all the trk files with the same name into their appropriate experiment directory. In terms of running tracker.track over multiple movies. I think you could write a loop to launch tracker.track for a single movie, wait until the trk file exists and then launch the next tracker.track. I think you may need to close the tracking monitor as well. This would at least require less manual interaction until we get the serialization online. Alice

…

On Wed, Feb 19, 2020 at 1:40 PM 2909ft ***@***.***> wrote: Dear Alice, Thank you for the reply! :) When I export the tracking results for my 6 videos, I get 6 different trk files (with different names, that is), but all from the same tracker type. What I was doing until now was editing the jaaba project before opening on training mode and importing the features from APT again, and repeating this process for each video that I wanted to add to that project. I was just wondering if there was a faster way of doing this, since this will take quite some time to do when the amount of videos I have increases substantially. Kind regards, Filipa A quarta, 19/02/2020, 18:44, arobie ***@***.***> escreveu: > Not sure I total understand. Do you have movies with trk files from > different trackers or just different trk file names from the same tracker? > > If its the first, I wouldn't recommend this because it might cause a > systematic difference in the perframe features for the different movies. > You may need to delete the previously calculated perframe features if you > change the trackersto force recalculation of the features. I think we > should add a check for this. > > If its just a naming difference, you can either make a new jab file that > points to the correct trk file name (you can import movies and labels from > old jab file). Or you can rename the trk files to match what JAABA expects. > JAABA is not setup to take trk files with different names for different > movies. > > Alice > > On Wed, Feb 19, 2020 at 11:40 AM 2909ft ***@***.***> wrote: > > > Also, another issue (this one more related to JAABA itself): I am trying > > to add new experiments to a jaaba classifier, but when I do so, JAABA > tries > > to use the original trk file (the one I initally used when creating the > > classifier) to generate the new features on the perframe directory. > > However, the new experiments have different trk files, which means that > > JAABA is not be able to find the original trk, giving me an error > message. > > Is there a way to get around this? > > > > — > > You are receiving this because you commented. > > Reply to this email directly, view it on GitHub > > < > #329?email_source=notifications&email_token=AAISKEHM2LNPGPECZJYMJSTRDVOHVA5CNFSM4KUPS6EKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEMIQBBQ#issuecomment-588316806 > >, > > or unsubscribe > > < > https://github.com/notifications/unsubscribe-auth/AAISKEGND4NDWZLKAG5VJL3RDVOHVANCNFSM4KUPS6EA > > > > . > > > > — > You are receiving this because you authored the thread. > Reply to this email directly, view it on GitHub > < #329?email_source=notifications&email_token=ANW7XOCF5T6VEVDFZVH2YDTRDVVWVA5CNFSM4KUPS6EKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEMIYHNA#issuecomment-588350388 >, > or unsubscribe > < https://github.com/notifications/unsubscribe-auth/ANW7XOCMN7ME5BX4NSRMGYTRDVVWVANCNFSM4KUPS6EA > > . > — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#329?email_source=notifications&email_token=AAISKEDUUL27KJXM7DTNRITRDV4KJA5CNFSM4KUPS6EKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEMI7F7Y#issuecomment-588378879>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAISKEF77D2V37AJ3CKSNITRDV4KJANCNFSM4KUPS6EA> .

[2909ft]

Dear Allen,

Yes, that would be great, thank you so much once again!

[2909ft]

Dear Alice,

Ah, those are 2 great ideas!! Thanks a lot!

[allenleetc]

Hi Filipa,

Thanks for your patience with this. We fixed bulk tracking for local backends (Docker, Conda). The wiki https://github.com/kristinbranson/APT/wiki/Tracking-Movies-(DL-algorithms) has a few updates. If you have a single GPU in your workstation, APT will track your movies serially.

Can you please give this a try? If you encounter issues, please post under kristinbranson/APT#263 if possible.

Thanks,
Allen

[2909ft]

Hi Allen!

Thank you very much for setting this up! I currently do not have access to a computer with a GPU since, due to the coronavirus situation, my workplace was closed without previous notice (I work in Italy). As soon as I do, I will test it :)

Thanks,
Filipa