Merge pull request #97 from lucasb-eyer/svhn

Add SVHN to the datasets.

Author: lucasb-eyer

DeepFried2/datasets/__init__.py

@@ -3,4 +3,5 @@
 from . import cifar100
 from . import sequence_sum
 from . import stanfordbg
+from . import svhn2
 from . import timeseries

DeepFried2/datasets/cifar10.py

@@ -18,21 +18,21 @@ def data():
                 batch = _pickle.load(b, encoding='latin1')
                 datas.append(_np.array(batch['data'], dtype=_np.float32))
                 labels.append(_np.array(batch['labels']))
-        Xtr = _np.concatenate(datas)
+        Xtr = _np.concatenate(datas).reshape((-1, 3, 32, 32))
         ytr = _np.concatenate(labels)
         Xtr /= 255
 
         # ... and the fifth as validation set as described in cuda-convnet:
         # https://code.google.com/p/cuda-convnet/wiki/Methodology
         with f.extractfile('cifar-10-batches-py/data_batch_5') as b:
             batch = _pickle.load(b, encoding='latin1')
-        Xva = _np.array(batch['data'], dtype=_np.float32)
+        Xva = _np.array(batch['data'], dtype=_np.float32).reshape((-1, 3, 32, 32))
         yva = _np.array(batch['labels'])
         Xva /= 255
 
         with f.extractfile('cifar-10-batches-py/test_batch') as b:
             batch = _pickle.load(b, encoding='latin1')
-        Xte = _np.array(batch['data'], dtype=_np.float32)
+        Xte = _np.array(batch['data'], dtype=_np.float32).reshape((-1, 3, 32, 32))
         yte = _np.array(batch['labels'])
         Xte /= 255
 

DeepFried2/datasets/cifar100.py

@@ -13,7 +13,7 @@ def data():
     with _taropen(fname, 'r') as f:
         with f.extractfile('cifar-100-python/train') as train:
             train = _pickle.load(train, encoding='latin1')
-        Xtr = _np.array(train['data'], dtype=_np.float32)
+        Xtr = _np.array(train['data'], dtype=_np.float32).reshape((-1, 3, 32, 32))
         ytr_c = _np.array(train['coarse_labels'])
         ytr_f = _np.array(train['fine_labels'])
         Xtr /= 255
@@ -27,7 +27,7 @@ def data():
 
         with f.extractfile('cifar-100-python/test') as test:
             test = _pickle.load(test, encoding='latin1')
-        Xte = _np.array(test['data'], dtype=_np.float32)
+        Xte = _np.array(test['data'], dtype=_np.float32).reshape((-1, 3, 32, 32))
         yte_c = _np.array(test['coarse_labels'])
         yte_f = _np.array(test['fine_labels'])
         Xte /= 255

DeepFried2/datasets/svhn2.py

@@ -0,0 +1,20 @@
+from DeepFried2.zoo.download import download as _download
+
+import numpy as _np
+import scipy.io as _sio
+
+
+def data():
+    fname_train = _download('http://ufldl.stanford.edu/housenumbers/train_32x32.mat')
+    fname_test = _download('http://ufldl.stanford.edu/housenumbers/test_32x32.mat')
+    fname_extra = _download('http://ufldl.stanford.edu/housenumbers/extra_32x32.mat')
+
+    def loadxy(fname):
+        mat = _sio.loadmat(fname)
+        X = mat['X'].transpose(3, 2, 0, 1).astype(_np.float32)
+        X /= 255
+        y = mat['y'][:,0].astype(_np.int64)  # For convenience/avoid stupid bugs.
+        return X, y
+
+    return loadxy(fname_train), loadxy(fname_test), loadxy(fname_extra)
+