I have created a KNN predictor for the MNIST dataset. The dataset contains images of handwritten images. The images are grayscale and measure 28 X 28 pixels. The information of these images is stored in 784 features, each feature corresponds to the intensity of a single pixel in the image. I want to implement this model to new images which are not in the test or training set. I want to write a piece of code which can take an image convert it into 28 X 28 pixels and then store the intensity of each of these pixels as a value in a unique feature. This will enable me to take new images and make predictions on them. What module should I use for this and what procedure should I follow?
You can use OpenCV and Numpy, Suppose you take a picture of a digit and save it as 'pic.jpg'
import cv2
import numpy as np
img = cv2.imread('pic.jpg', 0) # 0 for greyscale
resized = cv2.resize(img, (28,28))
features = resized.reshape(-1) #converts into single dimension
The features array will have the 784 features