How do I resize images in python for machine learning?

Question

I used a dataset of images for machine learning training. Each image had a width of 64px and a height of 64px as well. Now, I want to test my machine learning model using images from google. The problem is that google images are larger than training images, and I want to resize them so that their height and width are 64px (just like the images in the training set). Is there any way to do this in python? I did find some methods, but all of them maintain aspect ratio. So, I am unable to achieve 64 by 64 size.

What library do you plan on using for Machine Learning? Pytorch has a very nice function: `torchvision.transforms.CenterCrop()` which should work well. — ccl, Feb 27 '20 at 18:50
Not sure what you are looking for here. Do you need to resize the whole image to given size, or you want to used a cropped section.? — shaivikochar, Feb 27 '20 at 18:54
It seems that by default PIL doesn't maintain the aspect ratio, so I'm not sure what the issue is. — AMC, Feb 27 '20 at 19:02
from PIL import Image image = Image.open('./dataset/image.jpeg') image= image.resize((64,64)) — Saad Farooq, Feb 27 '20 at 19:02
Does this answer your question? [Does Python PIL resize maintain the aspect ratio?](https://stackoverflow.com/questions/2232742/does-python-pil-resize-maintain-the-aspect-ratio) — AMC, Feb 27 '20 at 19:07
Also: https://stackoverflow.com/questions/9983263/how-to-crop-an-image-using-pil — AMC, Feb 27 '20 at 19:07

score 1 · Answer 1 · edited Feb 27 '20 at 20:44

1

I found the function in PIL that resizes the image and does not maintain the aspect ratio.

from PIL import Image
image = Image.open('./dataset/image.jpeg')
image= image.resize((64,64))

edited Feb 27 '20 at 20:44

shaivikochar

400
2
7

answered Feb 27 '20 at 19:04

Saad Farooq

39
1
8

score 0 · Answer 2 · answered Feb 27 '20 at 19:09

You can use python-resize-image.

Install package:

pip install python-resize-image

Example:

from PIL import Image
from resizeimage import resizeimage

#open image file
with open('image.jpg', 'r+b') as fd_img:
    # create a PIL Image from file
    img = Image.open(fd_img)
    # resize image (contain)
    img = resizeimage.resize_contain(img, [64, 64])
    # covert to RBA incase it's RGBA
    img = img.convert("RGB")
    # save image
    img.save('resized-image.jpg', img.format)

How do I resize images in python for machine learning?

2 Answers2