Using HoG (Histogram with orientated gradient) and/or SIFT (Scale Invarient Feature Transform) together with the Bag of Words approach in an attempt to clasify images. In the case of Kaggle IET machine learning competition, where classifcation of 18 different image types(clothing) are needed, HoG with SIFT BOW has achieved the best accuracy (60%). This percentage was obtained after fine tuning using RandomizedSearchGrid. After the competition, i realized there may be better state-of-the art classifiers implemneted in Keras.