VOT Toolkit 5.0.1

[erhan gundogdu]

Dear VOT Committee,


Platform: Ubuntu 14.04, Matlab R2015a

Before the last update (in 5.0.0), everything was fine regarding the trax protocol.
Now, the problem is that if the tracker fails within a sequence, the tracker is not able to re-initialize itself. I debug it by writing some text in particular lines of my wrapper.

My guess is that the trax time out could not be set to the value that I want to, after the track failure since my tracker has a heavy initialization part.

What else could be the problem?

The error is given below:

      Tracker execution interrupted: Did not receive response.
Error using traxclient
Did not receive response.

Error in tracker_run (line 76)
    data = traxclient(tracker.command, callback, ...

Error in experiment_supervised (line 67)
    data = tracker_run(tracker, @callback, data);

Error in tracker_evaluate (line 57)
    [files, metadata] = experiment_function(tracker, sequence, directory,
        parameters, scan);

Error in workspace_evaluate>execute_iterator (line 127)
        tracker_evaluate(event.tracker, event.sequence, event.experiment);

Error in iterate (line 65)
            context = iterator(event, context);

Error in workspace_evaluate (line 93)
context = iterate(experiments, trackers, sequences, 'iterator', iterator, 'context',
context);

Error in run_experiments (line 11)
workspace_evaluate(tracker, sequences, experiments);
 

[Luka Čehovin Zajc]

From this output it is not entirely clear if there is a timeout, but in that case it is possible to increase the value. It may be best to enable TraX client debug output ( set_global_variable('trax_debug', true) ) and paste that here ... this way the problem may be more clear.

[erhan gundogdu]

Where should set_global_variable('trax_debug', true) be placed?
To the tracker_<<trackername>>.m or somewhere else?

I pasted it to tracker_<<trackername>>.m and run_experiments but there is no change in the error output.

[Luka Čehovin Zajc]

Usually it is placed in configuration.m, but I guess that tracker file should also work so I am surprised that it does not. One other option is also to enable console output instead of file using set_global_variable('trax_debug_console', true) ...

[erhan gundogdu]

Hi,

What I observed is that after a tracking failure [image, region] = traxserver('wait') command outputs an empty image path.
I realized this after putting a debug text file with an if statement.

The trax messages are given below and it says that there is a timeout reached:

@@TRAX:state "563.0000,347.0000,33.0000,11.0000"
CLIENT: Resetting program.
CLIENT: Trying to stop process nicely.
CLIENT: Flushing streams
CLIENT: Process should be terminated.
CLIENT: Stopping logger.
CLIENT: Tracker exited (stopped by signal 15)
CLIENT: Starting process /usr/local/MATLAB/R2015a/bin/matlab -nodesktop -nosplash -r "try; diary 'runtime.log'; addpath('/media/avaapm01/Erhan/VOT2017/CFCF_CCOT/VOT_integration/benchmark_wrapper');addpath('/media/avaapm01/Erhan/VOT2017/vot-toolkit/native/mex');CFCF_Wrapper_F; catch ex; disp(getReport(ex)); quit; end;"
CLIENT: Working directory is /tmp/tp5970f502_7241_4217_9052_8868cd240f29
@@TRAX:quit
CLIENT: Setting up TraX with standard streams connection
MATLAB is selecting SOFTWARE OPENGL rendering.

                            < M A T L A B (R) >
                  Copyright 1984-2015 The MathWorks, Inc.
                   R2015a (8.5.0.197613) 64-bit (glnxa64)
                             February 12, 2015

Warning: Duplicate directory name: /usr/local/MATLAB/R2015a/toolbox/local
 
To get started, type one of these: helpwin, helpdesk, or demo.
For product information, visit www.mathworks.com.
 
@@TRAX:hello "trax.image=path;" "trax.region=polygon;" "trax.version=1"
CLIENT: Tracker process ID: 29642
CLIENT: Connection with tracker established.
@@TRAX:initialize "file:///media/avaapm01/Erhan/VOT2017/vot-workspace/sequences/birds1/00000031.jpg" "546.9900,338.6700,599.2100,336.6500,600.6300,373.2800,548.4100,375.3000"
CLIENT: Timeout reached. Stopping tracker process ...
CLIENT: Trying to stop process nicely.
CLIENT: Flushing streams

CLIENT: Process should be terminated.
CLIENT: Stopping logger.
CLIENT: Tracker exited normally.
CLIENT: Cleaning up.
CLIENT: Stopping logger thread
CLIENT: Stopping watchdog thread

      Tracker execution interrupted: Did not receive response.
Error using traxclient
Did not receive response.

Error in tracker_run (line 76)
    data = traxclient(tracker.command, callback, ...

Error in experiment_supervised (line 67)
    data = tracker_run(tracker, @callback, data);

Error in tracker_evaluate (line 57)
    [files, metadata] = experiment_function(tracker, sequence, directory, parameters, scan);

Error in workspace_evaluate>execute_iterator (line 127)
        tracker_evaluate(event.tracker, event.sequence, event.experiment);

Error in iterate (line 65)
            context = iterator(event, context);

Error in workspace_evaluate (line 93)
context = iterate(experiments, trackers, sequences, 'iterator', iterator, 'context', context);

Error in run_experiments (line 12)
workspace_evaluate(tracker, sequences, experiments);

[Luka Čehovin Zajc]

Hm, this is indeed strange. I an currently getting ready to run some Matlab trackers on a similar setup so hopefully I will encounter the same problem, then I can debug it. Does this happen consistently on every sequence?

[erhan gundogdu]

It only happens when there is a track failure.
If there is no track failure, the wrapper and toolkits run successfully.
I will try more sequences to catch an exception if exists.

Thanks.

[erhan gundogdu]

I think it is not a problem related to time out.
If I use a faster tracker which has a short initialization duration, the problem again occurs.

Also, the time between giving the error and the track failure is exactly the same as the duration I set for trax_timeout.
Hence, i think, the problem might be the trax server fails to get the image path after a failure and waits for the timeout period ??

[Luka Čehovin Zajc]

Apparently there is a problem with recent changes that I have so far not been able to diagnose. It happens only on Linux, and if you use a bit older version of TraX library then the problem disappears. So a temporary solution is to download this commit unzip it to native/trax and re-compile native files.

[erhan gundogdu]

The reason for changing to the latest VotToolkit was to repeat the real time experiments.

If the problem is temporary, I prefer to wait for you to fix in the latest version.

[Luka Čehovin Zajc]

Well you can use the new toolkit and a bit older library, that is no problem. But fortunately I have now fixed the bug, it was not timeout issue, but a strange system-specific problem that the new tracker process received some messages left for the previous process. It should work now if you delete native directory and let the toolkit download and recompile everything from github.

[erhan gundogdu]

Thank you for your answer.
The problem seems to be solved since I have not experienced an error.