Exploración inicial de los datos

In [12]:

    

import os
import random

from tqdm import tqdm 

import matplotlib.pyplot as plt
from skimage import io #Para leer las imágenes
from skimage import img_as_float

%matplotlib inline


    

In [3]:

    

ROOT_DIR = os.getcwd()
DATA_DIR = os.path.join(ROOT_DIR, 'german-traffic-signs')
TRAIN_DIR = os.path.join(DATA_DIR, 'training-set')
TEST_DIR = os.path.join(DATA_DIR, 'test-set')


    

In [4]:

    

os.path.split(ROOT_DIR)


    

    
Out[4]:





('C:\\Users\\srodri16\\Desktop', 'supervised-avanzado-german-traffic-signs')

In [5]:

    

assert os.path.exists(DATA_DIR)
assert os.path.exists(TRAIN_DIR)
assert os.path.exists(TEST_DIR)


    

In [6]:

    

def plot_images_per_class(path): 
    lista  = [(int(folder), len(os.listdir(os.path.join(path,folder))) - 1) for folder in os.listdir(path)]
    lista = sorted(lista, key = lambda x: x[0])
    xVals = [clase for clase, val in lista]
    yVals = [val for clase, val in lista]
    fig = plt.figure(figsize=[15,5])
    ax = fig.add_subplot(111)
    plt.bar(xVals,yVals,width = 0.5)
    ax.set_title('Number of pictures per class', color = 'k', alpha = 0.6)
    ax.set_xlabel('Class', fontsize = 14)
    ax.set_ylabel('Number of pictures', fontsize = 14)
    ax.set_axisbelow(True)


    

In [7]:

    

plot_images_per_class(TRAIN_DIR)


    

In [8]:

    

plot_images_per_class(TEST_DIR)


    

In [9]:

    

def plot_random_sample(path,n_fos = 5,n_im = 3, scale = True, values = False):
    hor = n_fos
    ver = min(n_fos,n_im)
    fos = random.sample(os.listdir(path), n_fos)
    lista = []
    for fo in tqdm(fos):
        lista.append([os.path.join(fo,fi) for fi in random.sample(os.listdir(os.path.join(path,fo)),n_im) if fi.endswith('.jpg')])
    new_style = {'grid': False}
    plt.rc('axes', **new_style)
    _, ax = plt.subplots(n_fos, n_im, sharex='col', sharey='row', figsize=(2*n_im, 2*n_fos))
    for i, files in enumerate(lista):
        for j, file in enumerate(files):
            if scale:
                img = img_as_float(io.imread(os.path.join(path,file)))
            else:
                img = io.imread(os.path.join(path,file))
            if values:
                print(img)
            ax[i, j].imshow(img)
            ax[i, j].set_title(file)
    plt.show()


    

In [11]:

    

plot_random_sample(TRAIN_DIR)


    

    




100%|███████████████████████████████████████████| 5/5 [00:00<00:00, 131.58it/s]






    

In [10]:

    

plot_random_sample(TEST_DIR)


    

    




100%|██████████████████████████████████████████| 5/5 [00:00<00:00, 1000.02it/s]






    

In [21]:

    

from sklearn.decomposition import PCA, IncrementalPCA
import numpy as np


    

In [15]:

    

def load_data(files, scale = True):
    X, y = [], []
    for file in tqdm(files):
        if file.endswith(".jpg"):
            y.append(int(os.path.split(os.path.split(file)[0])[-1]))
            if scale:
                X.append(img_as_float(io.imread(file)))
            else:
                X.append(io.imread(file))
            
    return np.array(X), np.array(y)

def get_files_path(path, ext = ".jpg", n = np.inf):
    file_paths = []
    for folder in tqdm(os.listdir(path)):
        files = os.listdir(os.path.join(path,folder))
        if len(files) <= n:
            file_paths += [os.path.join(path,folder,file) for file in files if file.endswith(ext)]
        else:
            file_paths += [os.path.join(path,folder, file) for file in random.sample(files,n) if file.endswith(ext)]
            
    return file_paths


    

In [16]:

    

train_files = get_files_path(TRAIN_DIR)


    

    




100%|██████████████████████████████████████████| 43/43 [00:12<00:00,  3.58it/s]

In [17]:

    

X, y = load_data(train_files)
print(X.shape)
print(y.shape)


    

    




100%|████████████████████████████████████| 39209/39209 [07:56<00:00, 82.34it/s]






    




(39209, 32, 32, 3)
(39209,)

In [19]:

    

X = np.reshape(X, newshape = (-1,32*32*3))
print(X.shape)


    

    




(39209, 3072)

In [22]:

    

ipca = IncrementalPCA()
ipca.fit(X)


    

    
Out[22]:





IncrementalPCA(batch_size=None, copy=True, n_components=None, whiten=False)

In [26]:

    

plt.step(range(1,len(ipca.explained_variance_ratio_)+1), np.cumsum(ipca.explained_variance_ratio_), where='mid',
         label='cumulative explained variance')
plt.bar(range(1,len(ipca.explained_variance_ratio_)+1), ipca.explained_variance_ratio_, alpha=0.9, align='center',
        label='individual explained variance')
plt.ylabel('Explained variance ratio')
plt.xlabel('Principal components')
plt.legend(loc='best')
plt.show()