0
$\begingroup$

I have a data set of 145 images that I divide in to 96 images for training and 49 others for test. I have three classes which are imbalanced in Training set as well as in test set. In training set: class one:16 images; class two:56 images and class three: 24 images. I used matlab to implement the multi-class (SVM) one against all, but i have a poor results in terms of Accurity mainly precission. I want to have a good classification so i can have a good measure of these images. Is there any one who can help?

$\endgroup$
  • 1
    $\begingroup$ Unless you have (1) very simple classifier such as linear model and (2) very few and simple features for each image, I would say that you need at least an order of magnitude or more training data. $\endgroup$ – Vladislavs Dovgalecs Mar 9 '18 at 20:47
  • $\begingroup$ my learning data is a matrix whose size is (96x15). So I have 15 features for each image. $\endgroup$ – dalila Mar 11 '18 at 9:28
  • $\begingroup$ With 95 examples and 15 features each, there is perhaps a chance to build a classifier. Have you tried a linear classifier as your baseline? $\endgroup$ – Vladislavs Dovgalecs Mar 12 '18 at 6:13
  • $\begingroup$ I have already tried the linear classifier but I have not the intended results as you can see. Class1 Class2 Class3 Accurity 0.7347 Precision 0.6667 0.7500 0.7500 Recall 0.6000 0.7778 0.7500 Specificity 0.9231 0.6818 0.9189 $\endgroup$ – dalila Mar 13 '18 at 18:36
  • $\begingroup$ I have already tried the linear classifier but I have not the intended results as you can see. Class1 Class2 Class3 Accurity 0.7347 Precision 0.6667 0.7500 0.7500 Recall 0.6000 0.7778 0.7500 Specificity 0.9231 0.6818 0.9189 $\endgroup$ – dalila Mar 13 '18 at 18:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.