Hello Himali Vaghela HoG feature is ok but you might also consider extracting some features using some transfer learning models before applying Decision Tree Classifier on them
hi, decision trees take feature vectors as input. So you have to extract the characters you want from the image according to the information you are looking for there are several methods: co-occurrence matrices, SIFT, HoG, DCT .... And these are the characteristics that you will send the decision tree