Merge branch 'svhn'

README.md

@@ -20,6 +20,8 @@ Hi there, and welcome to the `extra-keras-datasets` module! This extension to th
   * [EMNIST-MNIST](#emnist-mnist)
   * [KMNIST-KMNIST](#kmnist-kmnist)
   * [KMNIST-K49](#kmnist-k49)
+  * [SVHN-Normal](#svhn-normal)
+  * [SVHN-Extra](#svhn-extra)
 - [Contributors and other references](#contributors-and-other-references)
 - [License](#license)
 
@@ -120,12 +122,36 @@ from extra-keras-datasets import kmnist
 
 ---
 
+### SVHN-Normal
+
+```
+from extra-keras-datasets import svhn
+(input_train, target_train), (input_test, target_test) = svhn.load_data(type='normal')
+```
+
+<a href="./assets/svhn-normal.png"><img src="./assets/svhn-normal.png" width="500" style="border: 3px solid #f6f8fa;" /></a>
+
+---
+
+### SVHN-Extra
+
+```
+from extra-keras-datasets import svhn
+(input_train, target_train), (input_test, target_test) = svhn.load_data(type='extra')
+```
+
+<a href="./assets/svhn-extra.png"><img src="./assets/svhn-extra.png" width="500" style="border: 3px solid #f6f8fa;" /></a>
+
+---
+
 ## Contributors and other references
 * **EMNIST dataset:**
   * Cohen, G., Afshar, S., Tapson, J., & van Schaik, A. (2017). EMNIST: an extension of MNIST to handwritten letters. Retrieved from http://arxiv.org/abs/1702.05373
   * [tlindbloom](https://stackoverflow.com/users/4008755/tlindbloom) on StackOverflow: [loading EMNIST-letters dataset](https://stackoverflow.com/questions/51125969/loading-emnist-letters-dataset/53547262#53547262) in [emnist.py](./emnist.py).
 * **KMNIST dataset:**
   * Clanuwat, T., Bober-Irizar, M., Kitamoto, A., Lamb, A., Yamamoto, K., & Ha, D. (2018). Deep learning for classical Japanese literature. arXiv preprint arXiv:1812.01718. Retrieved from https://arxiv.org/abs/1812.01718
+* **SVHN dataset:**
+  * Netzer, Y., Wang, T., Coates, A., Bissacco, A., Wu, B., & Ng, A. Y. (2011). Reading digits in natural images with unsupervised feature learning. Retrieved from http://ufldl.stanford.edu/housenumbers/nips2011_housenumbers.pdf / http://ufldl.stanford.edu/housenumbers/
 
 ## License
 The licenseable parts of this repository are licensed under a [MIT License](./LICENSE), so you're free to use this repo in your machine learning projects / blogs / exercises, and so on. Happy engineering! 🚀

extra_keras_datasets/__init__.py

@@ -1,4 +1,5 @@
 from __future__ import absolute_import
 
 from . import emnist
-from . import kmnist
+from . import kmnist
+from . import svhn

extra_keras_datasets/svhn.py

@@ -0,0 +1,59 @@
+'''
+  Import the SVHN dataset
+  Source: http://ufldl.stanford.edu/housenumbers/
+  Description: Street View House Numbers
+
+  ~~~ Important note ~~~
+  Please cite the following paper when using or referencing the dataset:
+  Yuval Netzer, Tao Wang, Adam Coates, Alessandro Bissacco, Bo Wu, Andrew Y. Ng Reading Digits in Natural Images with Unsupervised Feature Learning NIPS Workshop on Deep Learning and Unsupervised Feature Learning 2011. Retrieved from http://ufldl.stanford.edu/housenumbers/nips2011_housenumbers.pdf
+
+'''
+
+from keras.utils.data_utils import get_file
+import numpy as np
+from zipfile import ZipFile
+from scipy import io as sio
+import os
+
+def load_data(path='svhn_matlab.npz', type='normal'):
+  """Loads the SVHN dataset.
+  # Arguments
+      path: path where to cache the dataset locally
+          (relative to ~/.keras/datasets).
+      type: any of normal, extra (extra appends ~530K extra images for training)
+  # Returns
+      Tuple of Numpy arrays: `(input_train, target_train), (input_test, target_test)`.
+  """
+  path_train = get_file(f'{path}_train',
+                    origin='http://ufldl.stanford.edu/housenumbers/train_32x32.mat')
+  path_test = get_file(f'{path}_test',
+                    origin='http://ufldl.stanford.edu/housenumbers/test_32x32.mat')
+
+  # Load data from Matlab file.
+  # Source: https://stackoverflow.com/a/53547262
+  mat_train = sio.loadmat(path_train)
+  mat_test = sio.loadmat(path_test)
+
+  # Prepare training data
+  input_train = mat_train['X']
+  input_train = np.rollaxis(input_train, 3, 0)
+  target_train = mat_train['y'].flatten()
+
+  # Prepare testing data
+  input_test = mat_test['X']
+  input_test = np.rollaxis(input_test, 3, 0)
+  target_test = mat_test['y'].flatten()
+
+  # Append extra data, if required
+  if type == 'extra':
+    path_extra = get_file(f'{path}_extra',
+                  origin='http://ufldl.stanford.edu/housenumbers/extra_32x32.mat')
+    mat_extra = sio.loadmat(path_extra)
+    input_extra = mat_extra['X']
+    input_extra = np.rollaxis(input_extra, 3, 0)
+    target_extra = mat_extra['y'].flatten()
+    input_train = np.concatenate((input_extra, input_train))
+    target_train = np.concatenate((target_extra, target_train))
+
+  # Return data
+  return (input_train, target_train), (input_test, target_test)