Open PetreanuAndi opened 7 years ago
I'm not able to replicate the error. I tried running it on a new machine with Ubuntu 14.04 and the demo runs fine.
I have the same problem; provided CPLEX crashes with Matlab 2016b and 2017a on Linux.
actually me too :-1:
cplexmiqcp.m with Licensed Materials , Copyright IBM Corporation 2008 , 2017
which I replace in the cplex folder , copyright 2008 - 2010 causes the crash report in matlab .
where only function prototype is ad try/catch block catches that exception in solveMBIP.m
[x, fval, exitflag, output] = cplexmilp(fE,A,b,Aeq,beq,[],[],[],zeros(size(fE)),[],ctype,x0,options);
Output argument "x" (and maybe other) not assigned during call to "solveMBIP"
Please try to download the CPLEX library from the website following the links below. If you are a student you can get a license for free. https://www.ibm.com/support/knowledgecenter/en/SSSA5P_12.6.2/ilog.odms.cplex.help/CPLEX/MATLAB/topics/gs_install.html
I've followed the installation instructions on a Ubuntu machine, and on a Mac.
1) Ubuntu : Ran the test twice,once on a weak-performance local machine and once a strong-performance AWS session, and it always comes down to this : (tested on two our of the three provided datasets)
"Segmentation Violation....." (I guess this is a classic seg-fault ) inside of the black box cplexmilp(.....) call. -----> Matlab crash ----> end of session.
2) Mac : After installing the trial CPLEX and replacing the files in the project cplex folder, I get this error : "CPLEX Error 1016: Promotional version. Problem size limits exceeded." After looking online for a solution to this "problem size / promotional version issue", it seems like even academic licenses (those advertised as free) have issues with higher degrees of complexity of the underlining graph-model.
I'm new to using any type of CPLEX/optimisation library, so please excuse my lack of relevant vocabulary. I'm simply trying to find the best tracking solution out there (one that takes into consideration occlusion on multiple-frames while still performing almost real-time for <100 individual detections/frame).
:( any thoughts on the matter?