Hellwalker / randomforest-matlab

Automatically exported from code.google.com/p/randomforest-matlab
0 stars 0 forks source link

Matlab crashes #50

Closed GoogleCodeExporter closed 8 years ago

GoogleCodeExporter commented 8 years ago
Hello Abhishek, 

First of all thank you very much for making your code publicly available for 
research. I am using an hierarchical object recognition model that I have 
created,  which reduces to vectors learned by a classifier. I use a SVM but I 
am experimenting on other classifiers. 

I tried therefore to apply Random forests, but in both regression and 
classification as soon as the code hits your MEX files (which I compiled 
successfully) it crashes to desktop. 

I am running Windows 7 and the Matlab version is 2011a at 64bit.

Many Thanks
Aris

Original issue reported on code.google.com by tsitiri...@gmail.com on 22 Nov 2012 at 4:51

GoogleCodeExporter commented 8 years ago
Hi Aris

are you using the code from the download or from the svn? Also what compiler is 
being used with matlab? just to make sure its not a compiler specific issue can 
you also try the precompiled version? 
http://code.google.com/p/randomforest-matlab/downloads/detail?name=Windows-Preco
mpiled-RF_MexStandalone-v0.02-.zip&can=2&q=#makechanges

regards

Original comment by abhirana on 22 Nov 2012 at 5:02

GoogleCodeExporter commented 8 years ago
Thanks for your reply. I used the code following the link you sent and not it 
seemed, the code was able to train but crashed at "classRF_predict" (testing) 
and here is why...(i figured it out)

I ran the demo and it ran successfully, so it made be look harder in my code 
and realised that I was passing the testing vectors in a column. That's because 
from classifier to classifier vectors are requested in different format so I 
overlooked that...Anyway, it seems that simply using the row (transpose) 
worked. Isn't odd though that with just a minor mistake like that it would 
crash to desktop?

For your information, my compiler is VS 2010. 

Original comment by tsitiri...@gmail.com on 23 Nov 2012 at 10:09

GoogleCodeExporter commented 8 years ago
the major part of the code uses mex files and i think i am not doing a through 
check of whether the sizes of data match or not. the C code in the mex file is 
very finicky if the sizes are incorrect and that causes the crash. maybe i 
should put some strict checks.

great that you got a solution to your problem. hope you find the package 
useful. do tell if you run into any issues.

Original comment by abhirana on 25 Nov 2012 at 9:28