add RandomApply in gluon's transforms

python/mxnet/gluon/data/vision/transforms.py

@@ -581,3 +581,28 @@ def hybrid_forward(self, F, x):
         if is_np_array():
             F = F.npx
         return F.image.random_lighting(x, self._alpha)
+
+class RandomApply(Sequential):
+    """Apply a list of transformations randomly given probability
+
+    Parameters
+    ----------
+    Inputs:
+        - **transforms**: list of transformations
+        - **p**: probability
+
+    Outputs:
+        Transformed image.
+    """
+
+    def __init__(self, transform, p=0.5):
+        super(RandomApply, self).__init__()
+        self.transform = transform
+        self.p = p
+
+    def forward(self, x):
+        import random
+        if self.p < random.random():
+            return x
+        x = self.transform(x)
+        return x

tests/python/unittest/test_gluon_data_vision.py

@@ -229,6 +229,21 @@ def test_transformer():
     transform(mx.nd.ones((245, 480, 3), dtype='uint8')).wait_to_read()
 
 
+@with_seed()
+def test_randomtransforms():
+    from mxnet.gluon.data.vision import transforms
+
+    transform = transforms.Compose([transforms.RandomApply(transforms.Compose([transforms.Resize(300), transforms.RandomResizedCrop(224)]), 0.5)])
+
+    img = mx.nd.ones((245, 480, 3), dtype='uint8')
+    iteration = 1000
+    num_apply = 0
+    for i in range(iteration):
+        out = transform(img)
+        if out.shape[0] == 224:
+            num_apply += 1
+    assert(abs(num_apply/float(iteration)-0.5) < 1e-1)
+
 
 if __name__ == '__main__':
     import nose