X
PrevPrev Go to previous topic
NextNext Go to next topic
Last Post 21 Apr 2010 03:11 AM by  anon
Problem with ENVI EX`s algorithm (SVM)
 3 Replies
Sort:
You are not authorized to post a reply.
Author Messages

anon



New Member


Posts:
New Member


--
21 Apr 2010 03:11 AM
    Normal 0 false false false EN-US JA AR-SA MicrosoftInternetExplorer4 Hi everyone I have problem with feature extraction. I want to perform to supervise classification. I used feature extraction in ENVI EX, I select support vector machine algorithm with Kernel type of Polynomial. However, my pc could not perform. Could you help me about this problem? Best Hadi Fadaei

    Deleted User



    New Member


    Posts:
    New Member


    --
    21 Apr 2010 03:59 PM
    Hi Hadi, Sorry to hear you are running into problems with EX. I'm not sure what you could be ecountering, more details would help. I'm guessing that you went through the Scale and Merge without any problems. The assumption also is that you chose No Thresholding, chose All Attributes (Spatial, Spectral, Texture). Then you probably did Choose by Selecting Examples, clicked the Algorithm tab to access SVM and set up some parameters. Did you leave things at default besides Kernel type? You would then have to chose at least 2 Features in order to allow the Next button to be active. Was the problem that you only chose 1 feature and couldn't go to the next step or did you chose more and something else happened? What exactly do you mean by 'my PC could not perform? Ciao

    Deleted User



    New Member


    Posts:
    New Member


    --
    21 Apr 2010 07:41 PM
    Normal 0 false false false EN-US JA AR-SA MicrosoftInternetExplorer4 Hi Ciao Exactly I chosed two features with default of SVM algorithm. In fact, I want to counting trees in the high-resolution image satellite of ALOS. I performed SVM algorithm by fixed defaults. Nevertheless, my pc just run and more thinking without any result. I do not know how I can perform. Thank you for your helping me. Best Hadi   

    Deleted User



    New Member


    Posts:
    New Member


    --
    22 Apr 2010 01:22 PM
    Hi Hadi, SVM is computionally quite intensive and if your image is large it could take many hours. If your image is large you might want to run the SVM over night. To test if it is working you could use a subset of the image and run your SVM on that. Try using a subset that is 500 x 500 and see what happens.  
    You are not authorized to post a reply.