Machine Learning - Potato Leaf Disease Prediction

Step 1:

import numpy as np
import pandas as pd
import splitfolders
import matplotlib.pyplot as plt
import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import models, layers

Step 2:

# splitfolders.ratio will create 3 folders named [ train, test, val ]
# ratio = ( .8, .1, .1 )
# move=True will move the image to the respective folder. 

splitfolders.ratio(".", output=".", seed=1337, ratio=(.8, .1,.1,), group_prefix=None, move=False) 

Step 3:

IMAGE_SIZE = 256
CHANNELS = 3

    Part1: 

from tensorflow.keras.preprocessing.image import ImageDataGenerator

train_datagen = ImageDataGenerator(
        rescale=1./255,
        rotation_range=10,
        horizontal_flip=True
)
train_generator = train_datagen.flow_from_directory(
        'train',
        target_size=(IMAGE_SIZE,IMAGE_SIZE),
        batch_size=32,
        class_mode="sparse",
)

    Part2:

validation_datagen = ImageDataGenerator(
        rescale=1./255,
        rotation_range=10,
        horizontal_flip=True)
validation_generator = validation_datagen.flow_from_directory(
        'val',
        target_size=(IMAGE_SIZE,IMAGE_SIZE),
        batch_size=32,
        class_mode="sparse"
)

    Part3: 

test_datagen = ImageDataGenerator(
        rescale=1./255,
        rotation_range=10,
        horizontal_flip=True)

test_generator = test_datagen.flow_from_directory(
        'test',
        target_size=(IMAGE_SIZE,IMAGE_SIZE),
        batch_size=32,
        class_mode="sparse"
)

Step 4:

shape = (IMAGE_SIZE, IMAGE_SIZE, CHANNELS)
n_classes = 3
from tensorflow.keras.models import Sequential

model = models.Sequential([   
    layers.Conv2D(32, kernel_size = (3,3), activation='relu'),
    layers.MaxPooling2D((2, 2)),
    layers.Conv2D(64,  kernel_size = (3,3), activation='relu'),
    layers.MaxPooling2D((2, 2)),
    layers.Conv2D(64,  kernel_size = (3,3), activation='relu'),
    layers.MaxPooling2D((2, 2)),
    layers.Conv2D(64, (3, 3), activation='relu'),
    layers.MaxPooling2D((2, 2)),
    layers.Conv2D(64, (3, 3), activation='relu'),
    layers.MaxPooling2D((2, 2)),
    layers.Conv2D(64, (3, 3), activation='relu'),
    layers.MaxPooling2D((2, 2)),
    layers.Flatten(),
    layers.Dense(64, activation='relu'),
    layers.Dense(n_classes, activation='softmax'),
])

model.compile(
    optimizer='adam',
    loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False),
    metrics=['accuracy']
)

Step 5:

from tensorflow.keras.callbacks import TensorBoard
tensorboard_callback = TensorBoard(log_dir='./logs')  # Define a log directory

history = model.fit(
    train_generator,
    batch_size=32,
    validation_data=validation_generator,
    epochs=20,
    callbacks=[tensorboard_callback]
)

Step 6:

model.summary()

Step 7:

model.save('potato.keras')