diff --git a/tensorflow_graphics/image/pyramid.py b/tensorflow_graphics/image/pyramid.py
index 86bf08a54..ce088fd0c 100644
--- a/tensorflow_graphics/image/pyramid.py
+++ b/tensorflow_graphics/image/pyramid.py
@@ -21,15 +21,19 @@
 from __future__ import division
 from __future__ import print_function
 
+from typing import Callable, List, Optional, Tuple
+
 import numpy as np
 from six.moves import range
 import tensorflow as tf
 
 from tensorflow_graphics.util import export_api
 from tensorflow_graphics.util import shape
+from tensorflow_graphics.util import type_alias
 
 
-def _downsample(image, kernel):
+def _downsample(image: type_alias.TensorLike,
+                kernel: type_alias.TensorLike) -> tf.Tensor:
   """Downsamples the image using a convolution with stride 2.
 
   Args:
@@ -48,7 +52,8 @@ def _downsample(image, kernel):
       input=image, filters=kernel, strides=[1, 2, 2, 1], padding="SAME")
 
 
-def _binomial_kernel(num_channels, dtype=tf.float32):
+def _binomial_kernel(num_channels: int,
+                     dtype: tf.DType = tf.float32) -> tf.Tensor:
   """Creates a 5x5 binomial kernel.
 
   Args:
@@ -65,7 +70,9 @@ def _binomial_kernel(num_channels, dtype=tf.float32):
   return tf.constant(kernel, dtype=dtype) * tf.eye(num_channels, dtype=dtype)
 
 
-def _build_pyramid(image, sampler, num_levels):
+def _build_pyramid(image: type_alias.TensorLike, sampler: Callable[
+    [type_alias.TensorLike, type_alias.TensorLike], type_alias.TensorLike],
+                   num_levels: int) -> List[tf.Tensor]:
   """Creates the different levels of the pyramid.
 
   Args:
@@ -87,7 +94,8 @@ def _build_pyramid(image, sampler, num_levels):
   return levels
 
 
-def _split(image, kernel):
+def _split(image: type_alias.TensorLike,
+           kernel: type_alias.TensorLike) -> Tuple[tf.Tensor, tf.Tensor]:
   """Splits the image into high and low frequencies.
 
   This is achieved by smoothing the input image and substracting the smoothed
@@ -112,7 +120,10 @@ def _split(image, kernel):
   return high, low
 
 
-def _upsample(image, kernel, output_shape=None):
+def _upsample(image: type_alias.TensorLike,
+              kernel: type_alias.TensorLike,
+              output_shape: Optional[type_alias.TensorLike] = None
+              ) -> tf.Tensor:
   """Upsamples the image using a transposed convolution with stride 2.
 
   Args:
@@ -139,7 +150,9 @@ def _upsample(image, kernel, output_shape=None):
       padding="SAME")
 
 
-def downsample(image, num_levels, name="pyramid_downsample"):
+def downsample(image: type_alias.TensorLike,
+               num_levels: int,
+               name: str = "pyramid_downsample") -> List[tf.Tensor]:
   """Generates the different levels of the pyramid (downsampling).
 
   Args:
@@ -165,7 +178,8 @@ def downsample(image, num_levels, name="pyramid_downsample"):
     return _build_pyramid(image, _downsample, num_levels)
 
 
-def merge(levels, name="pyramid_merge"):
+def merge(levels: List[type_alias.TensorLike],
+          name: str = "pyramid_merge") -> tf.Tensor:
   """Merges the different levels of the pyramid back to an image.
 
   Args:
@@ -196,7 +210,9 @@ def merge(levels, name="pyramid_merge"):
     return image
 
 
-def split(image, num_levels, name="pyramid_split"):
+def split(image: type_alias.TensorLike,
+          num_levels: int,
+          name: str = "pyramid_split") -> List[tf.Tensor]:
   """Generates the different levels of the pyramid.
 
   Args:
@@ -228,7 +244,9 @@ def split(image, num_levels, name="pyramid_split"):
     return levels
 
 
-def upsample(image, num_levels, name="pyramid_upsample"):
+def upsample(image: type_alias.TensorLike,
+             num_levels: int,
+             name: str = "pyramid_upsample") -> List[tf.Tensor]:
   """Generates the different levels of the pyramid (upsampling).
 
   Args: