todo list:
(1) descriptors generation
(2) kmeans clustering to generate vocabulary (mean of each cluster) (3) generate histogramme for each image in the base, save the histogrammes
(4) given a new test image, calculate hist, compare it to all the hists in the base
(5) get the ranking of closest hist
(6) changes parameters and run tests

Guide to use:

(1)first, run

python visualWordsGen.py

this code will generate descpritors for the train dataset (if descpts_addr = '') and use them to train a kmeans model (parameters can be
configurated in line 51), then the kmeans models (~ 1 hr for training) will be saved for further usage.

(2) second, test for a target image, run

python test.py

this code will generate histogram (global descriptors) for each image from trainset (if hist_addr = '', or it will load
from the existed hists previously saved (the pretrained kmeans models will be loaded and used).
Then, we just calculate the distance (default: euclidean distance), return the top 10 nearest neighbors in terms of distances from the target (index and image name).