GNetTrainer - Application for training and predicting classification models.

Official Repository for GNetTrainer.

You can find the Package here- https://pypi.org/project/GNetTrainer/

Author

• Rishav Dash
• Mail me at - [email protected]
• Github - https://github.com/Rishav-hub/

About GNetTrainer

GNetTrainer is a Deep Learning web application for training and predicting classification models which is written in Python 3. At the backend it uses Frameworks like Tensorflow 2.0 and Keras. You can train any classification model with any data using GNetTrainer without writing any line of code.

Motivation

The main aim is to make something like Keras which is a high-level neural network library that runs on top of TensorFlow.

How Image Classification looks using Keras

Import the required libraries

from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Conv2D, MaxPool2D, Flatten, Dense, Dropout
from tensorflow.keras import optimizers
from tensorflow.keras.applications.densenet import DenseNet169, preprocess_input, decode_predictions
import numpy as np
import matplotlib.pyplot as plt
from PIL import Image
import cv2

Define Paths

ROOT = 'H:\\Parsonal\\Coding Practice\\dogCat'
os.chdir(ROOT)
os.getcwd()

Apply Augmentation

from keras.preprocessing.image import ImageDataGenerator
train_datagen = ImageDataGenerator(rescale = 1./255, 
                                   shear_range = 0.2,
                                   zoom_range = 0.2,
                                   horizontal_flip = True)
test_datagen = ImageDataGenerator(rescale = 1./255)


TRAIN_DIR = '/content/flowers_filtered/train'
TEST_DIR = '/content/flowers_filtered/val'

training_set = train_datagen.flow_from_directory(TRAIN_DIR, 
                                                 target_size = (224, 224),
                                                 batch_size = 32,
                                                 class_mode = 'categorical'
                                                 )

test_set = test_datagen.flow_from_directory(TEST_DIR, 
                                                 target_size = (224, 224),
                                                 batch_size = 32,
                                                 class_mode = 'categorical'
                                                 )
                                        

Download the pretrained model

model = DenseNet169(include_top=False, weights='imagenet', input_shape=(224, 224, 3))

Freeze layers

from keras.layers import BatchNormalization
for layer in model_base.layers:
    if isinstance(layer, BatchNormalization):
        layer.trainable = True
    else:
        layer.trainable = False

Add Custom Layers

from keras.layers import BatchNormalization
for layer in model_base.layers:
    if isinstance(layer, BatchNormalization):
        layer.trainable = True
    else:
        layer.trainable = False

Adding Custom Layers

model = Sequential()

model.add(tf.keras.layers.experimental.preprocessing.Resizing(224, 
                        224, interpolation="bilinear")) 

model.add(model_base)

model.add(Flatten())
model.add(Dense(128, activation='elu'))
model.add(Dropout(0.5))
model.add(BatchNormalization())
model.add(Dense(64, activation='elu'))
model.add(Dropout(0.5))
model.add(BatchNormalization())

model.add(Dense(5, activation='softmax'))

Defining Optimizers, Loss Functions and Checkpoints

OPTIMIZERS = optimizers.Adam()

checkpoints = tf.keras.callbacks.ModelCheckpoint(
    'Densnet_model_best.hdf5',
    monitor="val_loss",
    verbose=0,
    save_best_only=False)


lr_scheduler = tf.keras.callbacks.ReduceLROnPlateau(factor=0.5, patience=5) 

model.compile(optimizer= OPTIMIZERS, loss='categorical_crossentropy', metrics=['acc'])

Defining Path to Sabe the Model

import time
import os

def saveModel_path(model_dir="/content/drive/MyDrive/DLCVNLP/Computer_Vision/SAVED_MODELS"):
    os.makedirs(model_dir, exist_ok=True)
    fileName = time.strftime("DensNetModel_%Y_%m_%d_%H_%M_%S_.h5")    
    model_path = os.path.join(model_dir, fileName)
    print(f"your model will be saved at the following location\n{model_path}")
    return model_path

Tensorboard Callback

log_dir = get_log_path()
tb_cb = tf.keras.callbacks.TensorBoard(log_dir=log_dir)

Compiling Model

model.compile(optimizer= OPTIMIZERS, loss='categorical_crossentropy', metrics=['acc'])

Fit Model and Start Training

model.fit(training_set, 
          steps_per_epoch= 3452 // 32, 
          epochs = 10,
          validation_data = test_set,
          validation_steps = 10,
          callbacks= [checkpoints, lr_scheduler])

Epoch 1/10
107/107 [==============================] - 112s 817ms/step - loss: 0.9201 - acc: 0.6760 - val_loss: 0.5511 - val_acc: 0.8313
Epoch 2/10
107/107 [==============================] - 80s 746ms/step - loss: 0.4234 - acc: 0.8681 - val_loss: 0.3199 - val_acc: 0.8906
Epoch 3/10
107/107 [==============================] - 80s 747ms/step - loss: 0.3139 - acc: 0.9056 - val_loss: 0.2199 - val_acc: 0.9281
Epoch 4/10
107/107 [==============================] - 80s 746ms/step - loss: 0.2258 - acc: 0.9330 - val_loss: 0.2631 - val_acc: 0.9062
Epoch 5/10
107/107 [==============================] - 80s 748ms/step - loss: 0.2054 - acc: 0.9336 - val_loss: 0.2168 - val_acc: 0.9219
Epoch 6/10
107/107 [==============================] - 80s 748ms/step - loss: 0.1848 - acc: 0.9424 - val_loss: 0.3004 - val_acc: 0.9094
Epoch 7/10
107/107 [==============================] - 81s 749ms/step - loss: 0.1687 - acc: 0.9515 - val_loss: 0.2496 - val_acc: 0.9312

Image Classification using GNetTrainer

Create conda environment

conda create -n GNetTrainer python=3.6

Activate conda environment

conda activate GNetTrainer

Now all set to install the GNetTrainer Package

pip install GNetTrainer

To start the Magic. In the terminal type

gnet

Demo

Home Page

Training Page

Prediction Page