Bag of words (BOW) model is used in natural language processing for document classification where the frequency of each word is used as a feature to train a classifier. This approach can be extended to classify images by treating each feature in the image as a word, bag of visual words (BOVW). For BOVW approach, the features from image can be extracted using one of many feature extractors such as SIFT, HOG, SURF, and ORB. The extracted features can then be clustered into various bins using a clustering algorithm such as k-means. Finally, the features can be used to train a classifier such as Random Forest or SVM.
This tutorial explains the Python approach for BOVW based image classification.
0 Comments