ImageSeeker 0.2

Its an auto image classification library

ImageSeeker: An Image Classification Library

This repository hosts the development of the ImageSeeker library.

Authors

• @Bappy Ahmed

Email: [e-mail]

Github: https://github.com/entbappy/

🚀 About Me

I'm a Data Science learner. This library I have implemented just for learning purpose.

About ImageSeeker

ImageSeeker is a deep learning image classification library written in Python, running on top of the machine learning platform TensorFlow.Keras. It was developed with a focus on enabling fast experimentation of images classification. You can classify any image with any classification model in Keras appliaction by just writing 4 lines of code.

ImageSeeker is:

• Simple -- ImageSeeker reduces developer time of writing too much code of any image classification problem so that they focus on the parts of the problem that really matter.
• Flexible -- Keras adopts the principle of progressive disclosure of complexity: simple workflows should be quick and easy that’s why ImageSeeker is Flexible.
• Powerful -- Keras provides industry-strength performance and scalability: so we can use ImageSeeker in the production

Keras & ImageSeeker

Keras is the high-level API of TensorFlow 2: an approachable, highly-productive interface for solving machine learning problems, with a focus on modern deep learning. We can do so many tasks using Keras and image classification is one of them, but the issue is like we need to write many lines of code to implement an image classification solution.

But in ImageSeeker you don’t need to write many lines of code for implementing an image classification solution. You don’t also have to worry about your data preparation. What you need to do is just have to define your data path & some of the parameters of the model yes, your work will be done!

Image Classification using Keras

Necessary library importing

from tensorflow.keras.applications.densenet import VGG16
import tensorflow.keras as keras
from tensorflow.keras import models
from tensorflow.keras import layers
from tensorflow.keras import optimizers
import tensorflow as tf
from keras.models import load_model
import numpy as np
import matplotlib.pyplot as plt
import os

Redirecting to the data path

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

Preparing data & applying 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_path = "H:\\Parsonal\\Coding Practice\\dogCat\\train"
Test_path = "H:\\Parsonal\\Coding Practice\\dogCat\\valid"

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


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

Dowloading models

VGG = tf.keras.applications.VGG16(
    include_top=False,
    weights="imagenet",
    input_shape=(224,224,3)

)

VGG.summary()

Freezing layers

for layer in VGG.layers:
    layer.trainable = False

Adding custom layers

model = models.Sequential()
model.add(VGG)               
model.add(layers.Flatten())
model.add(layers.Dense(128, activation='relu' ))  
model.add(layers.Dropout(0.2))
model.add(layers.Dense(2, activation='softmax')) 
#model.summary()

Defining optimizers and loss function

from tensorflow.keras.optimizers import RMSprop
loss = 'categorical_crossentropy'
optimizer =RMSprop(learning_rate=0.0001)

Defining Tensorboard log and checkpoint

# Log

import time 

def get_log_path(log_dir="logs/fit"):
  fileName = time.strftime("log_%Y_%m_%d_%H_%M_%S")
  logs_path = os.path.join(log_dir, fileName)
  print(f"Saving logs at {logs_path}")
  return logs_path

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


# checkpoint

CKPT_path = "Model_ckpt.h5"
checkpointing_cb = tf.keras.callbacks.ModelCheckpoint(CKPT_path, save_best_only=True)

Compiling model

model.compile(optimizer = optimizer, loss=loss, metrics=['accuracy'])

Start training

hist = model.fit(training_set,
                         steps_per_epoch = 10,
                         epochs = 5,
                         validation_data = test_set,    
                         validation_steps = 20,
                         callbacks=[tb_cb,checkpointing_cb]
                         )

Epoch 1/5 10/10 [==============================] - 231s 21s/step - loss: 3.4007 - accuracy: 0.3812 - val_loss: 0.8920 - val_accuracy: 0.6438

Epoch 2/5 10/10 [==============================] - 100s 10s/step - loss: 1.0077 - accuracy: 0.6375 - val_loss: 0.6927 - val_accuracy: 0.7328

Epoch 3/5 10/10 [==============================] - 83s 9s/step - loss: 0.6629 - accuracy: 0.7344 - val_loss: 0.4763 - val_accuracy: 0.8266

Epoch 4/5 10/10 [==============================] - 69s 7s/step - loss: 0.4356 - accuracy: 0.8406 - val_loss: 0.6980 - val_accuracy: 0.7859

Epoch 5/5 10/10 [==============================] - 64s 6s/step - loss: 0.7226 - accuracy: 0.7844 - val_loss: 0.3199 - val_accuracy: 0.8906

Note: Then again you need write code for prediction as well

Image Classification using ImageSeeker

Define some parameters in the configuration.py

from tensorflow.keras.applications import VGG16

# Configure your data

TRAIN_DATA_DIR = "H:\\Parsonal\\Coding Practice\\iNeuron\\Moduler Coding\\ImageSeeker\\data\\train"        # Your training data path
VALID_DATA_DIR = "H:\\Parsonal\\Coding Practice\\iNeuron\\Moduler Coding\\ImageSeeker\\data\\valid"       # Your validation data path
CLASSES = 2                                                                                               # Number of classes in your data
IMAGE_SIZE = (224,224,3)                                                                                  #Image resulution/dimention with respect to your classification models
AUGMENTATION = True                                                                                       # If you want to apply Augmentation in your data (Default is True)
BATCH_SIZE = 32                                                                                           # Number of batch  (Default is 32)
PREDICTION_DATA_DIR = 'H:\\Parsonal\\\Coding Practice\\iNeuron\\Moduler Coding\\ImageSeeker\\prediction'  # Your prediction/test data path


# Configure your model

MODEL_OBJ = ResNet50(include_top=False,weights="imagenet",input_shape=(224,224,3))                         # Your pretrain model object  
PRETRAIN_MODEL_DIR = None                                                                                  #If you have any pretrain model exist path (Default is None)
MODEL_NAME ='ResNet50'                                                                                     # Your model name
EPOCHS = 5                                                                                                 # Number of Epochs
OPTIMIZER = 'adam'                                                                                         # Optimizers name/object
LOSS_FUNC = 'categorical_crossentropy'                                                                     # Your loss function name/object
FREEZE_ALL= True                                                                                           # Model layers freezing (Default is True)
FREEZE_TILL=None                                                                                          # You can define number of freezing layers (Defualt is None)

Great! Now open your Notebook

# Import some Modules of ImageSeeker

import os
from utils import config
from utils import data_manager as dm
import train_engine
import predict_engine

Now reload the configuration you made change

config.configureData()
config.configureModel()

Start training

train_engine.train()

Detected pretrain model!! Model has been saved following directory : Models\VGG16.h5 Preparing model... Freezing all... Adding sigmoid... Model loaded!! Augmetation applied! Found 200 images belonging to 2 classes. Found 110 images belonging to 2 classes. Saving logs at Tensorboard/logs/fit\log_2021_09_09_02_24_11

Epoch 1/5 6/6 [==============================] - 35s 6s/step - loss: 6.3008 - accuracy: 0.5385 - val_loss: 1.5761 - val_accuracy: 0.5104

Epoch 2/5 6/6 [==============================] - 29s 5s/step - loss: 2.8433 - accuracy: 0.5109 - val_loss: 2.4668 - val_accuracy: 0.5312

Epoch 3/5 6/6 [==============================] - 29s 5s/step - loss: 2.3428 - accuracy: 0.4800 - val_loss: 0.8409 - val_accuracy: 0.5417

Epoch 4/5 6/6 [==============================] - 29s 5s/step - loss: 0.8796 - accuracy: 0.5927 - val_loss: 0.6251 - val_accuracy: 0.6667

Epoch 5/5 6/6 [==============================] - 29s 5s/step - loss: 0.6290 - accuracy: 0.6746 - val_loss: 0.5840 - val_accuracy: 0.7188

Model saved at the following location : New_trained_model/newVGG16.h5

Yes thats it! You are done!!!

See the classes

dm.class_name()

Augmetation applied!

Found 200 images belonging to 2 classes.

Found 110 images belonging to 2 classes.

{'cat': 0, 'dog': 1}

Now start predicting

predict_engine.predict()

Original image : cat.2012.jpg. Predicted as [0]

Original image : cat.2018.jpg. Predicted as [0]

Original image : dog.2001.jpg. Predicted as [1]

Original image : dog.2002.jpg. Predicted as [1]

If you want to see the Tensorboard log

%load_ext tensorboard
%tensorboard --logdir Tensorboard/logs/fit

There you go!!! 😀 Your work is Done!!

Installation / Setup of ImageSeeker

You have to just clone ImageSeeker and install reqiurements.txt

 git clone https://github.com/entbappy/ImageSeeker-An-Image-Classification-Library.git

Now locate your terminal to ImageSeeker directory

 pip install -r reqiurements.txt

After cloning the folder structure should look like that...

Directory structure -

├── [data]             ## This folder contains your train & valid data
│    └────[train]
│    │       ├── [cat]
│    │       └── [dog]
     └────[valid]
             ├── [cat]
             ├── [dog]
├── [utils]
│   ├── __init__.py
│   ├── callbacks.py
│   └── config.py
    ├── data_manager.py
│   └── model.py
     
├── [Checkpoint]       ## Training checkpoints will automatically create here
│    │   
│    └── [models]
│        ├── VGG16.h5
│        └── Resnet50.h5  
|── [Models]           ## Pretrain model will be saved here
|── [New_trained_model] ## Training model will be saved here
|── [Prediction]
          ├── dog.jpg  ## testing image for prediction
│         └── cat.jpeg ## testing image for prediction
├── [Tensorboard]      ## Tensorboard log will be generated here
├── Tutorial.ipynb
├── configuration.py
├── train_engine.py
|── predict_engine.py
├── README.md
├── requirements.txt

Acknowledgements

• Keras
• SUNNY BHAVEEN CHANDRA
• Python