celeba_viz.ipynb

Kernel: Python 3

Visualize CelebA

Here we download a zipfile of images and their attributes that have been preprocessed to 64x64 using the script at https://github.com/probml/pyprobml/blob/master/scripts/celeba_kaggle_preprocess.py

# Standard Python libraries
from __future__ import absolute_import, division, print_function, unicode_literals

import os
import time
import numpy as np
import glob
import matplotlib.pyplot as plt
import PIL
import imageio
from IPython import display
import sklearn

# from time import time

np.random.seed(0)

# N can be 200, 20000, or 40000
N = 20000
H = 64
W = 64
C = 3
input_shape = [H, W, 3]
name = "celeba_small_H{}_W{}_N{}".format(H, W, N)
csv_name = "{}.csv".format(name)
zip_name = "{}.zip".format(name)

!rm {csv_name}
!wget https://raw.githubusercontent.com/probml/pyprobml/master/data/CelebA/{csv_name}

rm: cannot remove 'celeba_small_H64_W64_N20000.csv': No such file or directory
--2020-12-10 06:22:02--  https://raw.githubusercontent.com/probml/pyprobml/master/data/CelebA/celeba_small_H64_W64_N20000.csv
Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 151.101.0.133, 151.101.64.133, 151.101.128.133, ...
Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|151.101.0.133|:443... connected.
HTTP request sent, awaiting response... 200 OK
Length: 2440084 (2.3M) [text/plain]
Saving to: ‘celeba_small_H64_W64_N20000.csv’

celeba_small_H64_W6 100%[===================>]   2.33M  --.-KB/s    in 0.1s    

2020-12-10 06:22:03 (16.2 MB/s) - ‘celeba_small_H64_W64_N20000.csv’ saved [2440084/2440084]

import pandas as pd

df = pd.read_csv(csv_name)
df.head()

!rm {zip_name}
!wget https://raw.githubusercontent.com/probml/pyprobml/master/data/CelebA/{zip_name}

rm: cannot remove 'celeba_small_H64_W64_N20000.zip': No such file or directory
--2020-12-10 06:22:20--  https://raw.githubusercontent.com/probml/pyprobml/master/data/CelebA/celeba_small_H64_W64_N20000.zip
Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 151.101.0.133, 151.101.64.133, 151.101.128.133, ...
Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|151.101.0.133|:443... connected.
HTTP request sent, awaiting response... 200 OK
Length: 34148268 (33M) [application/zip]
Saving to: ‘celeba_small_H64_W64_N20000.zip’

celeba_small_H64_W6 100%[===================>]  32.57M  58.6MB/s    in 0.6s    

2020-12-10 06:22:24 (58.6 MB/s) - ‘celeba_small_H64_W64_N20000.zip’ saved [34148268/34148268]

!rm *.jpg
!ls

rm: cannot remove '*.jpg': No such file or directory
celeba_small_H64_W64_N20000.csv  celeba_small_H64_W64_N20000.zip  sample_data

!unzip -qq {zip_name}

from glob import glob

filenames = glob("*.jpg")
# print(filenames) # should match df['image_id']
print(len(filenames))

20000

from matplotlib.image import imread

images_celeba = np.zeros((N, H, W, C), dtype=np.float32)  # pre-allocate memory
for i in range(N):
    filename = df.iloc[i]["image_id"]
    img = imread(filename)  # numpy array of uint8
    images_celeba[i, :, :, :] = img / 255  # float in 0..1

fig, axs = plt.subplots(2, 4, figsize=(15, 10))
axs = np.reshape(axs, 8)
for i in range(8):
    ax = axs[i]
    ax.imshow(images_celeba[i, :, :, :])
    ax.axis("off")
plt.tight_layout()
plt.show()