dl-assistant 0.0.6

A libirary to automate developing keras models

PyPi Library :- dl-assistant Documentation

Table of Contents

• Installation
• Usage - Image Classification
• Examples
• API Reference

Installation

It can be easily installed by typing the following in the terminal:

pip install dl-assistant

Usage - Image classification

1. create_dataframe- This function takes dir containing training or test data and convert it into pandas dataframe that is 1st step towards deep learning

   Simple Usage

                           from dl_assistant.image_cnn
                           import classification
                           x=classification()
                           df=x.create_dataframe('train') # It will create a dataframe getting a datframe with image address in one column and label in another
                           #Continue on your own
   

2. prep_x_train- This function takes image paths, height and width to resize and returns an array of features of training data. It itself divide features by 255.0

   Simple Usage

                           from dl_assistant.image_cnn import classification
                           x=classification()
                           df=x.create_dataframe('train') # It will create a dataframe getting a datframe with image address in one column and label in another
                           x_train = x.prep_x_train(df['Images',100,100])# get features divided by 255.0 
   

3. prep_y_train- This function takes labels and number of classes as input and converts labels into binary metrics.

   Simple Usage

                           from dl_assistant.image_cnn import classification
                           x=classification()
                           df=x.create_dataframe('train') # It will create a dataframe getting a datframe with image address in one column and label in another
                           x_train = x.prep_x_train(df['Images',100,100])
                           y_train = x.prep_y_train(df['Labels'],7)
   

4. make_model- This function takes unit of 1 stconv2D layer and number of classes as input and returns a basic non-trained model.

   Simple Usage

                           from dl_assistant.image_cnn import classification
                           x=classification()
                           df=x.create_dataframe('train') # It will create a dataframe getting a datframe with image address in one column and label in another
                           x_train = x.prep_x_train(df['Images',100,100])
                           y_train = x.prep_y_train(df['Labels'],7)
                           shape=x_train[0].shape
                           model = x.make_model(128,7,shape)
   

5. expert_make_model- This is a function developed for expert.

   • Syntax =expert_make_model(self,layers,unit,num_classes,input_shape,dr)

     • layers= This is a string list that can contain either Conv2D,MaxPooling2D or Dropout
     • unit= This is the integer that contains the unit of 1st Conv2D layer
     • num_classes= This is the count of no. of labels
     • input_shape= This is the input shape or shape of x_train[0]
     • dr= The dropout rate

   Simple Usage

                           from dl_assistant.image_cnn import classification
                           x=classification()
                           df=x.create_dataframe('train') # It will create a dataframe getting a datframe with image address in one column and label in another
                           x_train = x.prep_x_train(df['Images',100,100])
                           y_train = x.prep_y_train(df['Labels'],7)
                           shape=x_train[0].shape
                           model = x.expert_make_model(['Conv2D','Conv2D','MaxPooling2D','Dropout'],128,7,shape,0.5)
   

6. train_model- This function takes model to train, x_train , y_train, epochs and batch_size and returns a trained model

   Simple Usage

                           from dl_assistant.image_cnn import classification
                           x=classification()
                           df=x.create_dataframe('train') # It will create a dataframe getting a datframe with image address in one column and label in another
                           x_train = x.prep_x_train(df['Images',100,100])
                           y_train = x.prep_y_train(df['Labels'],7)
                           shape=x_train[0].shape
                           model = x.make_model(128,7,shape)
                           batch_size=32
                           epochs=70
                           model.x.train_model(model,x_train,y_train,batch_size,epochs)
   

Examples

Create a emotion-classification model

            `from dl_assistant.image_cnn import classification                      model = classification()                     model = model.create('TRAIN',7)                     model.predict([x_test[0]])`