Every ResNet in TensorFlow 2

This repository contains implementations of every ResNet model descibed in the Deep Residual Learning for Image Recognition paper by Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun (2015). The models are implemented in the resnets.py file. The only requirement is TensorFlow 2 or higher (tested with TensorFlow 2.17.0).

The implemented models are:

• ResNet-18
• ResNet-34
• ResNet-50
• ResNet-101
• ResNet-152

Each model is implemented as a function that follows this signature:

def ResNetN(
    input_shape: Tuple[int, int, int],
    output_units: int = 1000,
    include_top: bool = True,
    after_input: Optional[Union[Sequential, Layer]] = None,
    normalize: bool = False,
    kernel_regularizer: Optional[Union[Regularizer, str]] = None,
    kernel_initializer: Union[Initializer, str] = "he_uniform",
    flatten: bool = False,
    dropout_rate: float = 0.0,
) -> Model

where N in the function name is the number of layers in the model that should be replaced with 18, 34, 50, 101, or 152. The function takes the following parameters:

• input_shape - Shape of the input images
• output_units - Number of output units used in the last layer if include_top is True (default: 1000)
• include_top - Whether to include the network top after global average pooling or the flatten layer (default: True)
• after_input - Custom layers to add after the input like preprocessing layers as a Keras model of class tf.keras.Sequential or as a single layer of class tf.keras.layers.Layer (default: None - no custom layers)
• normalize - Whether to normalize the input images to the range [0, 1] (default: False)
• kernel_regularizer - Kernel regularizer of class tf.keras.regularizers.Regularizer or as a string (default: None)
• kernel_initializer - Kernel initializer of class tf.keras.initializers.Initializer or as a string (default: "he_uniform")
• flatten - Whether to use a flatten layer instead of a global average pooling layer after the last block (default: False - use global average pooling)
• dropout_rate - Dropout rate used after global average pooling or flattening (default: 0.0)

Models use the Functional API of Keras under the hood which defines the model's structure as a directed acyclic graph of layers. The function returns a tf.keras.Model instance that needs to be compiled and trained.

The implementation was tested by successfully creating each model and printing its summary to text files. A helper function to print model summaries to text files was used:

def write_summary(model: Model, file_path: str) -> None:
    with open(file_path, "w") as f:
        model.summary(print_fn=lambda x: f.write(x + "\n"))

The text files are included in the repository and are named resnet18.txt, resnet34.txt, resnet50.txt, resnet101.txt, and resnet152.txt.