update

ace19-dev · Aug 8, 2019 · ae5b213 · ae5b213
1 parent 165e8f3
commit ae5b213
Show file tree

Hide file tree

Showing 2 changed files with 10 additions and 4 deletions.
diff --git a/retrieval/retrieval_data.py b/retrieval/retrieval_data.py
@@ -58,7 +58,6 @@ def decode(self, serialized_example):
         # Convert from a scalar string tensor to a float32 tensor with shape
         image_decoded = tf.image.decode_png(features['image/encoded'], channels=3)
         image = tf.image.convert_image_dtype(image_decoded, dtype=tf.float32)
-        image = tf.image.resize(image, [self.resize_h, self.resize_w])
 
         # Convert label from a scalar uint8 tensor to an int32 scalar.
         label = tf.cast(features['image/class/label'], tf.int64)
@@ -83,8 +82,12 @@ def eval(self, filename, image, label):
     def tencrop(self, filename, image, label):
         """Prepare one image for TenCrop
         """
-        flip_mode = random.randint(0, 1)
+        # Resize the image to the height and width.
+        image = tf.expand_dims(image, 0)
+        image = tf.image.resize(image, [self.resize_h, self.resize_w])
+        image = tf.squeeze(image, [0])
 
+        flip_mode = random.randint(0, 1)
         images = []
         for i in range(5):
             image = tf.random_crop(image, [RANDOM_CROP_SIZE, RANDOM_CROP_SIZE, 3])

diff --git a/val_data.py b/val_data.py
@@ -58,7 +58,6 @@ def decode(self, serialized_example):
         # Convert from a scalar string tensor to a float32 tensor with shape
         image_decoded = tf.image.decode_png(features['image/encoded'], channels=3)
         image = tf.image.convert_image_dtype(image_decoded, dtype=tf.float32)
-        image = tf.image.resize(image, [self.resize_h, self.resize_w])
 
         # Convert label from a scalar uint8 tensor to an int32 scalar.
         label = tf.cast(features['image/class/label'], tf.int64)
@@ -83,8 +82,12 @@ def eval(self, filename, image, label):
     def tencrop(self, filename, image, label):
         """Prepare one image for TenCrop
         """
-        flip_mode = random.randint(0, 1)
+        # Resize the image to the height and width.
+        image = tf.expand_dims(image, 0)
+        image = tf.image.resize(image, [self.resize_h, self.resize_w])
+        image = tf.squeeze(image, [0])
 
+        flip_mode = random.randint(0, 1)
         images = []
         for i in range(5):
             image = tf.random_crop(image, [RANDOM_CROP_SIZE, RANDOM_CROP_SIZE, 3])