Image Utility

Images are a huge part of neural networks. Classification of images is a huge resource for many projects requiring some form of computer vision. EasyNN has created a image class to make converting your image into the proper form to match the dataset your using to classify. Let's get to some code.

Import to get all the functionality.

from EasyNN.utilities.image.preprocess import image

Simple full example

from EasyNN.examples.mnist.number.trained import model
from EasyNN.utilities import Preprocess, download

# Download an example image.
download("four.jpg","https://bit.ly/3lAJrMe")

format_options = dict(
    grayscale=True,
    invert=True,
    process=True,
    contrast=30,
    resize=(28, 28),
    rotate=3,
)

# Converting your image into the correct format for the mnist number dataset.
image = Preprocess("four.jpg").format(**format_options)

# Show function is tied to a model.
model.show(image)

Image Parameters

These are the current attributes that the image class allows you to modify.

grayscale=True / False

The grayscale parameter will change the image into grayscale pixels that have a rand from 0-255 of intensity.

invert=True / False

Inverted the colors of the image. If the pixel is white its now black for grayscale values.

process=True / False

Preprocess allows for noise to be removed from your image by taking the average of all the pixels and remove up to that threshold.

Example of the pixels:

Average is: 207/9 = 23

Before          After:
[30,10,30],     [7,0,7],
[25,25,25],     [3,3,2],
[25,10,30]      [2,0,7]

Image example:

SHOW COMPARE IMAGES

contrast= 0-255

Contrast will remove noise but with an amount. The higher the number the more the image will remove pixel values.

resize=(28, 28) - (height,width) of pixels.

Change the image to a desired height and width.

rotate=1-4

Rotate the image. Number represents rotation by 90 degrees clockwise.