pyradox

This python library helps you with implementing various state of the art neural networks in a totally customizable fashion using Tensorflow 2

Installation

pip install pyradox

or

pip install git+https://github.com/Ritvik19/pyradox.git

Usage

Modules

Module | Description | Input Shape | Output Shape | Usage ---|---|---|---|--- Rescale | A layer that rescales the input: x_out = (x_in -mu) / sigma | Arbitrary | Same shape as input | check here Convolution 2D | Applies 2D Convolution followed by Batch Normalization (optional) and Dropout (optional) | 4D tensor with shape (batch_shape, rows, cols, channels) | 4D tensor with shape (batch_shape, new_rows, new_cols, new_channels) | check here Densely Connected | Densely Connected Layer followed by Batch Normalization (optional) and Dropout (optional) | 2D tensor with shape (batch_size, input_dim) | 2D tensor with shape (batch_size, n_units) | check here DenseNet Convolution Block | A Convolution block for DenseNets | 4D tensor with shape (batch_shape, rows, cols, channels) | 4D tensor with shape (batch_shape, new_rows, new_cols, new_channels) | check here DenseNet Convolution Block | A Convolution block for DenseNets |4D tensor with shape (batch_shape, rows, cols, channels) | 4D tensor with shape (batch_shape, new_rows, new_cols, new_channels) | check here DenseNet Transition Block | A Transition block for DenseNets | 4D tensor with shape (batch_shape, rows, cols, channels) | 4D tensor with shape (batch_shape, new_rows, new_cols, new_channels) | check here Dense Skip Connection | Implementation of a skip connection for densely connected layer | 2D tensor with shape (batch_size, input_dim) | 2D tensor with shape (batch_size, n_units) | check here VGG Module | Implementation of VGG Modules with slight modifications, Applies multiple 2D Convolution followed by Batch Normalization (optional), Dropout (optional) and MaxPooling | 4D tensor with shape (batch_shape, rows, cols, channels) | 4D tensor with shape (batch_shape, new_rows, new_cols, new_channels) | check here Inception Conv | Implementation of 2D Convolution Layer for Inception Net, Convolution Layer followed by Batch Normalization, Activation and optional Dropout | 4D tensor with shape (batch_shape, rows, cols, channels) | 4D tensor with shape (batch_shape, new_rows, new_cols, new_channels) | check here Inception Block | Implementation on Inception Mixing Block | 4D tensor with shape (batch_shape, rows, cols, channels) | 4D tensor with shape (batch_shape, new_rows, new_cols, new_channels) | check here Xception Block | A customised implementation of Xception Block (Depthwise Separable Convolutions) | 4D tensor with shape (batch_shape, rows, cols, channels) | 4D tensor with shape (batch_shape, new_rows, new_cols, new_channels) | check here Efficient Net Block | Implementation of Efficient Net Block (Depthwise Separable Convolutions) | 4D tensor with shape (batch_shape, rows, cols, channels) | 4D tensor with shape (batch_shape, new_rows, new_cols, new_channels) | check here Conv Skip Connection | Implementation of Skip Connection for Convolution Layer | 4D tensor with shape (batch_shape, rows, cols, channels) | 4D tensor with shape (batch_shape, new_rows, new_cols, new_channels) | check here Res Net Block | Customized Implementation of ResNet Block | 4D tensor with shape (batch_shape, rows, cols, channels) | 4D tensor with shape (batch_shape, new_rows, new_cols, new_channels) | check here Res Net V2 Block | Customized Implementation of ResNetV2 Block | 4D tensor with shape (batch_shape, rows, cols, channels) | 4D tensor with shape (batch_shape, new_rows, new_cols, new_channels) | check here Res NeXt Block | Customized Implementation of ResNeXt Block | 4D tensor with shape (batch_shape, rows, cols, channels) | 4D tensor with shape (batch_shape, new_rows, new_cols, new_channels) | check here Inception Res Net Conv 2D | Implementation of Convolution Layer for Inception Res Net: Convolution2d followed by Batch Norm | 4D tensor with shape (batch_shape, rows, cols, channels) | 4D tensor with shape (batch_shape, new_rows, new_cols, new_channels) | check here Inception Res Net Block | Implementation of Inception-ResNet block | 4D tensor with shape (batch_shape, rows, cols, channels) | 4D tensor with shape (batch_shape, new_rows, new_cols, new_channels) | Block 8 Block 17 Block 35 NAS Net Separable Conv Block | Adds 2 blocks of Separable Conv Batch Norm | 4D tensor with shape (batch_shape, rows, cols, channels) | 4D tensor with shape (batch_shape, new_rows, new_cols, new_channels) | check here NAS Net Adjust Block | Adjusts the input previous path to match the shape of the input | | | NAS Net Normal A Cell | Normal cell for NASNet-A | | | NAS Net Reduction A Cell | Reduction cell for NASNet-A | | | Mobile Net Conv Block | Adds an initial convolution layer with batch normalization and activation | 4D tensor with shape (batch_shape, rows, cols, channels) | 4D tensor with shape (batch_shape, new_rows, new_cols, new_channels) | check here Mobile Net Depth Wise Conv Block | Adds a depthwise convolution block. A depthwise convolution block consists of a depthwise conv, batch normalization, activation, pointwise convolution, batch normalization and activation | 4D tensor with shape (batch_shape, rows, cols, channels) | 4D tensor with shape (batch_shape, new_rows, new_cols, new_channels) | check here Inverted Res Block | Adds an Inverted ResNet block | 4D tensor with shape (batch_shape, rows, cols, channels) | 4D tensor with shape (batch_shape, new_rows, new_cols, new_channels) | check here SEBlock | Adds a Squeeze Excite Block | 4D tensor with shape (batch_shape, rows, cols, channels) | 4D tensor with shape (batch_shape, new_rows, new_cols, new_channels) | check here

ConvNets

Module | Description | Input Shape | Output Shape | Usage ---|---|---|---|--- Generalized Dense Nets | A generalization of Densely Connected Convolutional Networks (Dense Nets) | 4D tensor with shape (batch_shape, rows, cols, channels) | 4D tensor with shape (batch_shape, new_rows, new_cols, new_channels) | check here Densely Connected Convolutional Network 121 | A modified implementation of Densely Connected Convolutional Network 121 | 4D tensor with shape (batch_shape, rows, cols, channels) | 4D tensor with shape (batch_shape, new_rows, new_cols, new_channels) | check here Densely Connected Convolutional Network 169 | A modified implementation of Densely Connected Convolutional Network 169 | 4D tensor with shape (batch_shape, rows, cols, channels) | 4D tensor with shape (batch_shape, new_rows, new_cols, new_channels) | check here Densely Connected Convolutional Network 201 | A modified implementation of Densely Connected Convolutional Network 201 | 4D tensor with shape (batch_shape, rows, cols, channels) | 4D tensor with shape (batch_shape, new_rows, new_col