How to load images from URL using Pytorch - deep-learning

Question

I want to load the images using Pytorch
I have a dataset of image_urls with its corresponding labels(offer_id are labels.)
Is there any efficient way of doing it in Pytorch?.

Answer 1

This should work if image url is a public URL using pillow, requests and torch together
from PIL import Image
import requests
from torchvision.io import read_image
import torchvision.transforms as transforms
url = "https://example.jpg"
image = Image.open(requests.get(url, stream=True).raw)
transform = transforms.Compose([
transforms.PILToTensor()])
torch_image = transform(image)

Answer 2

You can use the requests package:
import requests
from PIL import Image
import io
import cv2
response = requests.get(df1.URL[0]).content
im = Image.open(io.BytesIO(response))

Answer 3

You may convert your image URLs to files first by downloading them to specific folder representing the label. You will certainly find a way to do so. Then you may do like this to check what you have:
%%time
import glob
f=glob.glob('/content/imgs/**/*.png')
print(len(f), f)
There is a need to create a image loader that will read the image from disk. In here the pil_loader:
def pil_loader(path):
with open(path, 'rb') as f:
img = Image.open(f)
return img.convert('RGB')
ds = torchvision.datasets.DatasetFolder('/content/imgs',
loader=pil_loader,
extensions=('.png'),
transform=t)
print(ds)
You may check how I did that for Cifar10.
Check the section "From PNGs to dataset".