Coverage for source/model/model_building_blocks/vgg16_block.py: 100%

1# model/model_building_blocks/vgg16_block.py 

2 

3import tensorflow as tf 

4from tensorflow.keras.layers import Conv2D, MaxPooling2D 

5 

6class Vgg16Block: 

7 """ 

8 Class implementing Vgg16 block compatible with tensorflow API. This block is a core component of  

9 the VGG16 architecture, applying two convolutional layers followed by a max pooling layer to  

10 downsample and extract features from the input tensor. 

11 

12 Diagram: 

13 

14 .. code-block:: text 

15 Input Tensor --> +-----------------------+ +-----------------------+ +-----------------------+  

16 | Conv2D | | Conv2D | | MaxPooling2D | 

17 | Filters: N1 |-->| Filters: N2 |-->| Pool Size: K3xK3 | 

18 | Kernel Size: K1xK1 | | Kernel Size: K2xK2 | | | 

19 +-----------------------+ +-----------------------+ +-----------------------+ --> Output Tensor  

20 """ 

21 

22 def __init__(self, kernels: tuple[tuple[int, int], tuple[int, int], tuple[int, int]], filters: tuple[int, int]) -> None: 

23 """ 

24 Class constructor. 

25 

26 Parameters: 

27 kernels (tuple[tuple[int, int], tuple[int, int], tuple[int, int]]): Sizes of all kernels used within this block. 

28 filters (tuple[int, int]): Number of filters used in convolutional layers.  

29 """ 

30 

31 self.__conv_2d_1_kernel_size: tuple[int, int] = kernels[0] 

32 self.__conv_2d_2_kernel_size: tuple[int, int] = kernels[1] 

33 self.__max_pooling_2d_kernel_size: tuple[int, int] = kernels[2] 

34 self.__conv_2d_1_nr_of_filters: int = filters[0] 

35 self.__conv_2d_2_nr_of_filters: int = filters[1] 

36 

37 def __call__(self, input_tensor: tf.Tensor) -> tf.Tensor: 

38 """ 

39 Applies convolutional transformation with max pooling to input tensor. 

40 

41 Parameters: 

42 input_tensor (tf.Tensor): Input tensor that transformations should be applied to. 

43 

44 Returns: 

45 (tf.Tensor): Output tensor with applied transformations. 

46 """ 

47 

48 x = Conv2D(self.__conv_2d_1_nr_of_filters, self.__conv_2d_1_kernel_size, 

49 activation = 'relu', padding = 'same')(input_tensor) 

50 x = Conv2D(self.__conv_2d_2_nr_of_filters, self.__conv_2d_2_kernel_size, 

51 activation = 'relu', padding = 'same')(x) 

52 

53 output_tensor = MaxPooling2D(self.__max_pooling_2d_kernel_size)(x) 

54 

55 return output_tensor