Data missmatch in pizza_steak_sushi (04_Custom_datasets)

[radzio-ui]

Hi,
I just found some wrong labeled data (randomly). I don't know if it matter to you right now, but still worth to mention. This is an image of burger inside pizza folder.
Course: 04_pytorch_custom_datasets
Timestamp: 20:56:57
Image: data\pizza_steak_sushi\test\pizza\2508636.jpg

Code (raw from training, just to let you know at which point I found it):

from pathlib import Path

import torch
from torch import nn
import torchvision
from torch.utils.data import DataLoader

device = 'cuda' if torch.cuda.is_available() else 'cpu'
data_path = Path("data/")
image_path = data_path / 'pizza_steak_sushi'


def download_data(image_path: Path, url):
    if image_path.is_dir():
        print(f"Image path {image_path} already exists")
    else:
        print(f"Image path {image_path} does not exist, creating path")
        image_path.mkdir(exist_ok=True, parents=True)

    import requests
    import zipfile
    request = requests.get(url)
    with open(str(image_path) + '.zip', 'wb') as f:
        print('Downloading data')
        f.write(request.content)
    with zipfile.ZipFile(image_path, 'r') as zip_file:
        print("Unzipping data", zip_file)
        zip_file.extractall(image_path.parent)


# import zipfile
# image_path = Path(r'C:\0_cpm_bev\pizza_steak_sushi.zip')
# with zipfile.ZipFile(image_path, 'r') as zip_file:
#     print("Unzipping data", zip_file)
#     zip_file.extractall(image_path.parent)

import os
def walk_through_dir(dir_path):
    for dirpath, dirnames, filenames in os.walk(dir_path):
        print(f' There are length {len(dirnames)} directories and {len(filenames)} images in "{dirpath}".')

# walk_through_dir(image_path)

train_dir = image_path / 'train'
test_dir = image_path / 'test'

from PIL import Image
import random
import matplotlib.pyplot as plt
import numpy as np
random.seed(42)

image_path_list = list(image_path.glob('*/*/*.jpg'))
# print(image_path_list)
random_image_path = random.choice(image_path_list)
print(random_image_path)

image_class = random_image_path.parent.stem
print(image_class)

image = Image.open(random_image_path)
print(f'Random image path: {random_image_path} of Image class {image_class}')
print(f'Image height: {image.height}, width: {image.width}')

def plot_images():
    cols = 4
    rows = 4
    fig = plt.figure(figsize=(9,9))
    for i in range(1, cols*rows+1):
        fig.add_subplot(cols, rows, i)
        image = random.choice(image_path_list)
        img = Image.open(image)
        plt.imshow(img)
        plt.title(image.parent.stem)
        plt.axis(False)
        print(f'{image}     shape {np.asarray(img).shape}')
    plt.show(block=False)

from torchvision import transforms, datasets

data_transform = transforms.Compose([
    transforms.Resize(size=(64,64)),
    transforms.RandomHorizontalFlip(p=0.5),
    transforms.ToTensor()
])
# data_transform = nn.Sequential( the same as above )
# print(data_transform(image).shape)

def plot_transformed_images(image_paths, transform, n=3, seed=None):
    if seed:
        random.seed(seed)
    random_image_paths  = random.sample(image_paths, n)
    for image_path in random_image_paths:
        print(image_path)
        with Image.open(image_path) as f:
            fig, ax = plt.subplots(nrows=1, ncols=2)
            ax[0].imshow(f)
            ax[0].set_title(f'Original\nsize: {f.size}')
            ax[0].axis(False)

            transformed_image = transform(f)
            ax[1].imshow(transformed_image.permute(1,2,0))
            ax[1].set_title(f'Transformed\nshape: {transformed_image.shape}')

            fig.suptitle(f'Class: {image_path.parent.stem}', fontsize=16)
    plt.show(block=False)
plot_transformed_images(image_path_list, data_transform)

[shisirkha]

This might be an issue with the dataset don't worry just remove the image and get another image from the internet