SVM IMPLEMENTATION

• One classifier is trained for each of the three classes. Each classifier is trained by considering the in-class data versus the rest of the data.

  
  

• The radial basis kernel function was chosen because it was found to give the best performance on a number of related classification tasks.

  
  

• For each test image, we calculate the distance of its feature vector to each of the three classifiers. Classification was achieved by comparing the distances to the three hyperplanes.