how to increase the accuracy of an image classifier? - tensorflow

Question

I made an image classifier using Tensorflow, Keras with the implementation of a CNN architecture, the model works pretty fine (at least for the images that I have tested on it ) and it has reached an accuracy of 78.87%, the only thing that I m facing is that I want to make the accuracy no less than 85%.
Please Note:
Dataset: 2 folders: [Train Folder===> 80 folders each has 110 images, Validation folder===> 80 folders each has 22 images] size of the images [240-260]x[40-60]
Below is the code I used to create, save and test my model:
from keras.preprocessing.image import ImageDataGenerator
from keras.models import Sequential
from keras.layers import Conv2D, MaxPooling2D
from keras.layers import Activation, Dropout, Flatten, Dense
from keras import backend as K
# dimensions of our images.
img_width, img_height = 251, 54
#img_width, img_height = 150, 33
train_data_dir = 'C:/Users/ADEM/Desktop/msi_youssef/PFE/test/numbers/data/train'
validation_data_dir = 'C:/Users/ADEM/Desktop/msi_youssef/PFE/test/numbers/data/valid'
nb_train_samples = 8800 #10435
nb_validation_samples = 1763 #2051
epochs = 30 #20 # how much time you want to train your model on the data
batch_size = 32 #16
if K.image_data_format() == 'channels_first':
input_shape = (3, img_width, img_height)
else:
input_shape = (img_width, img_height, 3)
model = Sequential()
model.add(Conv2D(32, (3, 3), input_shape=input_shape))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Conv2D(32, (3, 3)))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Conv2D(64, (3, 3)))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Flatten())
model.add(Dense(64))
model.add(Activation('relu'))
model.add(Dropout(0.5))
model.add(Dense(80)) #1
model.add(Activation('softmax')) #sigmoid
model.compile(loss='sparse_categorical_crossentropy',optimizer='rmsprop',metrics=['accuracy'])#categorical_crossentropy #binary_crossentropy
# this is the augmentation configuration we will use for training
train_datagen = ImageDataGenerator(
rescale=1. / 255,
shear_range=0.1,
zoom_range=0.05,
horizontal_flip=False)
# this is the augmentation configuration we will use for testing:
# only rescaling
test_datagen = ImageDataGenerator(rescale=1. / 255)
train_generator = train_datagen.flow_from_directory(
train_data_dir,
target_size=(img_width, img_height),
batch_size=batch_size,
class_mode='binary')
validation_generator = test_datagen.flow_from_directory(
validation_data_dir,
target_size=(img_width, img_height),
batch_size=batch_size,
class_mode='binary')
model.fit_generator(
train_generator,
steps_per_epoch=nb_train_samples // batch_size,
epochs=epochs,
validation_data=validation_generator,
validation_steps=nb_validation_samples // batch_size)
model.save('testX_2.h5') #first_try
last epoche resulat
Epoch 30/30
275/275 [==============================] - 38s 137ms/step - loss: 0.9406 - acc: 0.7562 - val_loss: 0.1268 - val_acc: 0.9688
how I tested my model:
from keras.models import load_model
from keras.preprocessing import image
import matplotlib.pyplot as plt
import numpy as np
import os
result = {"0":"0", "1":"0.25", "2":"0.5", "3":"0.75", "4":"1", "5":"1.25", "6":"1.5", "7":"1.75",
"47":"2", "48":"2.25", "49":"2.5", "50":"2.75", "52":"3","53":"3.25", "54":"3.5", "55":"3.75", "56":"4", "57":"4.25", "58":"4.5",
"59":"4.75","60":"5", "61":"5.25", "62":"5.5", "63":"5.75", "64":"6", "65":"6.25","66":"6.5", "67":"6.75", "68":"7", "69":"7.25",
"70":"7.5", "71":"7.75", "72":"8", "73":"8.25", "74":"8.5", "75":"8.75", "76":"9", "77":"9.25", "78":"9.5", "79":"9.75", "8":"10",
"9":"10.25", "10":"10.5", "11":"10.75", "12":"11", "13":"11.25", "14":"11.5", "15":"11.75", "16":"12","17":"12.25", "18":"12.5",
"19":"12.75", "20":"13", "21":"13.25", "22":"13.5", "23":"13.75","24":"14", "25":"14.25", "26":"14.5", "27":"14.75", "28":"15",
"29":"15.25", "30":"15.5", "31":"15.75", "32":"16", "33":"16.25", "34":"16.5", "35":"16.75", "36":"17", "37":"17.25", "38":"17.5",
"39":"17.75", "40":"18", "41":"18.25", "42":"18.5", "43":"18.75", "44":"19", "45":"19.25", "46":"19.5", "51":"20"}
def load_image(img_path, show=False):
img = image.load_img(img_path, target_size=(251, 54))
img_tensor = image.img_to_array(img) # (height, width, channels)
img_tensor = np.expand_dims(img_tensor, axis=0) # (1, height, width, channels), add a dimension because the model expects this shape: (batch_size, height, width, channels)
img_tensor /= 255. # imshow expects values in the range [0, 1]
if show:
plt.imshow(img_tensor[0])
plt.axis('off')
plt.show()
return img_tensor
if __name__ == "__main__":
# load model
model = load_model('C:/Users/ADEM/Desktop/msi_youssef/PFE/other_shit/testX_2.h5')
# image path
img_path = 'C:/Users/ADEM/Desktop/msi_youssef/PFE/dataset/5.75/a.png'
# load a single image
new_image = load_image(img_path)
# check prediction
#pred = model.predict(new_image)
pred = model.predict_classes(new_image)
#print(pred[0])
print(result[str(pred[0])])

Answer 1

Taking all the information about dataset and considering your CNN model already has around 80% accuracy you can start with training the model for a higher number of epochs (typically > 100 epochs). That should give the required boost to your model.
If that alone does not work you can implement:
Transformation/augmentation:
perform transformation/augmentation on the images before feeding into the model.
Tweak Model:
make changes to model layers and do hyperparameter tuning.
You can follow this article to learn more.