Implementing DTW

In this example, we will look for similarity in 684 images from 8 categories. We will use four imports of PIL, numpy, mlpy, and collections:

from PIL import Image
from numpy import array
import mlpy
from collections import OrderedDict

First, we need to obtain the time series representation of the images and store it in a dictionary (data) with the number of the image and its time series as data[fn] = list.

Tip

The performance of this process will lie in the number of images processed, so beware of the use of memory with large datasets.

data = {}

for fn in range(1,685):
  img = Image.open("ImgFolder\\{0}.jpg".format(fn))
  arr = array(img)
  list = []
  for n in arr: list.append(n[0][0])
  for n in arr: list.append(n[0][1])
  for n in arr: list.append(n[0][2])
  data[fn] = list

Then, we need to select an image for the reference, which will be compared with all the other images in the data dictionary:

reference = data[31]

Now, we need to apply the mlpy.dtw_std function to all the...