Flower Recognition Using Convolutional Neural Network

In this article, we will use Convolutional Neural Network to solve a Supervised Image 

Classification problem of recognizing the flower types – rose, chamomile, dandelion, sunflower, & tulip. Understanding the concept of a Convolutional Neural Network refers to this.

Once, understand the concept of CNN we can start with the project.

Steps to be followed 

The steps followed in this article are listed here :

1. Importing modules
2. Importing Dataset 
3. Image Data Generator
4. Model Development 
5. Model Evaluation and Prediction

Importing modules

• Pandas – This library helps to load the data frame in a 2D array format.
• Numpy – Numpy arrays are very fast and can perform large computations.
• Matplotlib – This library is used to draw visualizations.
• Sklearn – This module contains multiple libraries having pre-implemented functions to perform tasks from data preprocessing to model development and evaluation.
• OpenCV – This library focuses on image processing and handling.
• Tensorflow – It has a range of functions to achieve complex functionalities with single lines of code.

Python3

import numpy as np
import pandas as pd
import cv2
import matplotlib.pyplot as plt
from PIL import Image
from tensorflow.keras import layers
from tensorflow.keras.preprocessing.image import ImageDataGenerator, load_img
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Conv2D, MaxPooling2D
from tensorflow.keras.layers import Activation, Dropout, Flatten, Dense
from tensorflow.keras.optimizers import Adam
import tensorflow as tf
import os

Importing Dataset and Preprocessing

You can download the dataset from the link.

Once download the data, image resize is needed to be done. To specify the size of the image use below code. 

Python3

base_dir = '/flowers/'

img_size = 224
batch = 64

Image Data Generator

For image data generator, follow the code given below.

Python3

# Create a data augmentor
train_datagen = ImageDataGenerator(rescale=1. / 255, shear_range=0.2,
                                   zoom_range=0.2, horizontal_flip=True,
                                   validation_split=0.2)

test_datagen = ImageDataGenerator(rescale=1. / 255,
                                  validation_split=0.2)

# Create datasets
train_datagen = train_datagen.flow_from_directory(base_dir,
                                                  target_size=(
                                                      img_size, img_size),
                                                  subset='training',
                                                  batch_size=batch)
test_datagen = test_datagen.flow_from_directory(base_dir,
                                                target_size=(
                                                    img_size, img_size),
                                                subset='validation',
                                                batch_size=batch)

Output : 

Found 3121 images belonging to 5 classes.
Found 776 images belonging to 5 classes.

Model Development 

From here we will use the we creating CNN model with the help of tensorflow library as it contains all the functionalities that one may need to define the architecture of a Convolutional Neural Network and train it on the data.

Python3

# # modelling starts using a CNN.

model = Sequential()
model.add(Conv2D(filters=64, kernel_size=(5, 5), padding='same',
                 activation='relu', input_shape=(224, 224, 3)))
model.add(MaxPooling2D(pool_size=(2, 2)))


model.add(Conv2D(filters=64, kernel_size=(3, 3),
                 padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2)))


model.add(Conv2D(filters=64, kernel_size=(3, 3),
                 padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2)))

model.add(Conv2D(filters=64, kernel_size=(3, 3),
                 padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2)))

model.add(Flatten())
model.add(Dense(512))
model.add(Activation('relu'))
model.add(Dense(5, activation="softmax"))

Summary of the model

Python3

model.summary()

Output : 

Python3

keras.utils.plot_model(
    model,
    show_shapes = True,
    show_dtype = True,
    show_layer_activations = True
)

Output:

Compiling 

Python3

model.compile(optimizer=tf.keras.optimizers.Adam(),
              loss='categorical_crossentropy', metrics=['accuracy'])

Fitting

Model Fitting will take some time.

Python3

epochs=30
model.fit(train_datagen,epochs=epochs,validation_data=test_datagen)

Output:

Saving and Loading model

For more information regarding saving and loading models refer this.

Python3

from tensorflow.keras.models import load_model
model.save('Model.h5')

# load model
savedModel=load_model('Model.h5')

Once you saved the model, you can use it anytime without training again and again.

Model Evaluation and Prediction

Evaluation includes the hyperparameter tuning if the model is not giving good results. You can also play with different parameters for better predictions.

Python3

train_datagen.class_indices

Output :

 {'daisy': 0, 'dandelion': 1, 'rose': 2, 'sunflower': 3, 'tulip': 4}

Python3

from keras.preprocessing import image

#Creating list for mapping
list_ = ['Daisy','Danelion','Rose','sunflower', 'tulip']

#Input image
test_image = image.load_img('img.jpg',target_size=(224,224))

#For show image
plt.imshow(test_image)
test_image = image.img_to_array(test_image)
test_image = np.expand_dims(test_image,axis=0)

# Result array
result = savedModel.predict(test_image)
print(result)

#Mapping result array with the main name list
i=0
for i in range(len(result[0])):
  if(result[0][i]==1):
    print(list_[i])
    break

Output:

Python3

#Input image
test_image = image.load_img('img2.jpg',target_size=(224,224))

#For show image
plt.imshow(test_image)
test_image = image.img_to_array(test_image)
test_image = np.expand_dims(test_image,axis=0)

# Result array
result = savedModel.predict(test_image)
print(result)

#Mapping result array with the main name list
i=0
for i in range(len(result[0])):
  if(result[0][i]==1):
    print(list_[i])
    break

Output:

Conclusion

We can improve the results by using Transfer Learning like Resnet50.