from 

tensorflow.compat.v1 import ConfigProto
from tensorflow.compat.v1 import InteractiveSession

config = ConfigProto()
config.gpu_options.per_process_gpu_memory_fraction = 0.5
config.gpu_options.allow_growth = True
session = InteractiveSession(config=config)

from tensorflow.keras.layers import Input, Lambda, Dense, Flatten

from tensorflow.keras.models import Model

from tensorflow.keras.applications.resnet50 import ResNet50

#from keras.applications.vgg16 import VGG16

from tensorflow.keras.applications.resnet50 import preprocess_input

from tensorflow.keras.preprocessing import image

from tensorflow.keras.preprocessing.image import ImageDataGenerator,load_img

from tensorflow.keras.models import Sequential

import numpy as np

from glob import glob

import PIL

import matplotlib.pyplot as plt

import tensorflow as tf

from google.colab import drive

drive.mount('/content/drive')

Mounted at /content/drive

# re-size all the images to this

IMAGE_SIZE = [224, 224]

train_path = '/content/drive/MyDrive/DYNOCARD_DATA/Train'

valid_path = '/content/drive/MyDrive/DYNOCARD_DATA/Test'

train_ds = tf.keras.preprocessing.image_dataset_from_directory(train_path)

Found 1050 files belonging to 15 classes.

class_names = train_ds.class_names

print(class_names)

['ANCHORED BUT NOT SET', 'ASPHALTENES IN PUMP', 'BAD TAIL BEARING', 'DEEP WELL', 'FLU

#plt.figure(figsize=(20,20))

#for images, labels in train_ds.take(1):

#for i in range(2):

 # ax = plt.subplot(2,2, i + 1)

 # plt.imshow(images[i].numpy().astype("uint8"))

  #plt.title(class_names[labels[i]])

  #plt.axis("off")

import tensorflow

resnet50 =tensorflow.keras.applications.ResNet152V2(input_shape=IMAGE_SIZE + [3], weights=

Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/re

234561536/234545216 [==============================] - 1s 0us/step

# don't train existing weights

for layer in resnet50.layers:

    layer.trainable = False

# useful for getting number of output classes

folders = glob('/content/drive/MyDrive/DYNOCARD_DATA/Train/*')

# our layers - you can add more if you want

x = Flatten()(resnet50.output)

prediction = Dense(len(folders), activation='softmax')(x)

# create a model object

model = Model(inputs=resnet50.input, outputs=prediction)

model.summary()

# tell the model what cost and optimization method to use

model.compile(

  loss='categorical_crossentropy',

  optimizer='adam',

  metrics=['accuracy']

# Use the Image Data Generator to import the images from the dataset

from tensorflow.keras.preprocessing.image import ImageDataGenerator

train_datagen = ImageDataGenerator(rescale = 1./255,

                                   shear_range = 0.1,

                                   zoom_range = 0.1,

                                   horizontal_flip = False)

test_datagen = ImageDataGenerator(rescale = 1./255)

# Make sure you provide the same target size as initialied for the image size

training_set = train_datagen.flow_from_directory('/content/drive/MyDrive/DYNOCARD_DATA/Tra
                                                 target_size = (224, 224),

                                                 batch_size = 32,

                                                 class_mode = 'categorical')

Found 1050 images belonging to 15 classes.

test_set = test_datagen.flow_from_directory('/content/drive/MyDrive/DYNOCARD_DATA/Test',

                                            target_size = (224, 224),

                                            batch_size = 32,

                                            class_mode = 'categorical')

Found 448 images belonging to 15 classes.

#fluidpound = list(train_path.glob('fluidpound/*'))

#print(fluidpound[0])

#PIL.Image.open(str(fluidpound[0]))

import matplotlib.pyplot as plt

# fit the model

# Run the cell. It will take some time to execute

r = model.fit_generator(

  training_set,

  validation_data=test_set,

  epochs=7,

  steps_per_epoch=len(training_set),

  validation_steps=len(test_set)

/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:8: UserWarning: `Model.f

Epoch 1/7

33/33 [==============================] - 920s 28s/step - loss: 0.6901 - accuracy: 0.8

33/33 [==============================] - 30s 927ms/step - loss: 0.0050 - accuracy: 0

33/33 [==============================] - 27s 824ms/step - loss: 0.0112 - accuracy: 0

33/33 [==============================] - 27s 833ms/step - loss: 0.0023 - accuracy: 0

33/33 [==============================] - 28s 838ms/step - loss: 8.5169e-04 - accuracy

33/33 [==============================] - 28s 838ms/step - loss: 5.7457e-04 - accuracy

33/33 [==============================] - 28s 840ms/step - loss: 1.3858e-04 - accuracy

import matplotlib.pyplot as plt
# plot the loss
plt.plot(r.history['loss'], label='train loss')
plt.plot(r.history['val_loss'], label='val loss')
plt.ylabel("Percentage",
               fontweight ='bold',fontsize=20, c = "r")
plt.xlabel("No of EPOCHS",
               fontweight ='bold', c = "green",fontsize=16)
plt.title("RESNET Model Losses", fontsize=22, fontweight = "bold", loc="center", c="yellow
plt.grid(color = 'black', linestyle = '--', linewidth = 0.35)
plt.legend(["Training Loss", "Validation Loss"], loc ="upper right")
plt.show()
#plt.savfig('LossVal_loss')
print('')
# plot the accuracy
plt.plot(r.history['accuracy'], label='train acc')
plt.plot(r.history['val_accuracy'], label='val acc')
plt.ylabel("Percentage",
               fontweight ='bold',fontsize=20, c = "r")
plt.xlabel("No of EPOCHS",
               fontweight ='bold', c = "green",fontsize=16)
plt.title("RESNET Model Accuracy", fontsize=22, fontweight = "bold", loc="center", c="yell
plt.grid(color = 'black', linestyle = '--', linewidth = 0.35)
plt.legend(["Training Accuracy", "Validation Accuracy"], loc ="lower right")
plt.show()