from google.colab import drive
drive.mount('/content/drive')Mounted at /content/drive
-
Business Problem
-
Data Understanding
-
Data Preparation
-
Modelling
-
Evaluation
The data set we will use in this project is TrashNet. It was prepared by Stanaford University students. The dataset consists of six different classes. There are a total of 2527 images in the dataset, including Glass, Paper, Cardboard, Plastic, Metal and Garbage.
Distribution of images:
- 501 glasses
- 594 paper
- 403 cardboard
- 482 plastic
- 410 metal
- 137 trash
The images were taken by placing them on a white board and using sunlight or room lighting. The images have a size of 512x384 pixels and 3 (RGB) channels.
# Import the necessary libraries
import cv2
import urllib
import itertools
import numpy as np
import pandas as pd
import seaborn as sns
import random, os, glob
from imutils import paths
import matplotlib.pyplot as plt
from sklearn.utils import shuffle
from urllib.request import urlopen
from sklearn.metrics import classification_report
from sklearn.metrics import confusion_matrix
# It is used to switch off warnings.
import warnings
warnings.filterwarnings('ignore')
# Libraries to be used for the model
import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.preprocessing import image
from tensorflow.keras.utils import to_categorical
from tensorflow.keras.callbacks import ModelCheckpoint, EarlyStopping
from tensorflow.keras.layers import Conv2D, Flatten, MaxPooling2D, Dense, Dropout, SpatialDropout2D
from tensorflow.keras.preprocessing.image import ImageDataGenerator, img_to_array, load_img, array_to_img!pip install imutilsRequirement already satisfied: imutils in /usr/local/lib/python3.10/dist-packages (0.5.4)
- Pairing the Google Colab Notebook with the Google Drive account.
- Upload the data set to Google Drive and show the address to the notebook
- Defining the function to read the data set.
# Keeping the path information of the images in this data set in the drive
dir_path = '/content/drive/MyDrive/Garbage-Classification'# Determining Target size and Label Label Values
target_size = (224, 224)
waste_labels = {'cardboard': 0, 'glass': 1, 'metal': 2, 'paper': 3, 'plastic': 4, 'trash': 5}def load_datasets(path):
# Reads the image from the directory of images and creates the tags.
# Parameters
# path: The directory where the images are located.
# return:
# x: List holding the class information to which the images belong.
x = []
labels = []
# It lists and sorts the images in the path we send.
image_paths = sorted(list(paths.list_images(path)))
for image_path in image_paths:
# The images in the specified path are read with the openCV library.
img = cv2.imread(image_path)
# The read images are rescaled according to the initially specified target_size.
img = cv2.resize(img, target_size)
# Scaled images are added to the x list.
x.append(img)
# When each path is separated by '/', the second to last element of the returned list represents the label.
label = image_path.split(os.path.sep)[-2]
# A label is created by taking the value value corresponding to the key value sent from the waste_labels dictionary, which contains the numeric value equivalents of the captured labels.
labels.append(waste_labels[label])
# The data set is randomly shuffled.
x, labels = shuffle(x, labels, random_state = 42)
# Size and class information is reported.
print(f"X size: {np.array(x).shape}")
print(f"Label class number: {len(np.unique(labels))} Number of observations: {len(labels)}")
return x, labelsx, labels = load_datasets(dir_path)X size: (2527, 224, 224, 3)
Label class number: 6 Number of observations: 2527
# Retention of image sizes
input_shape = (np.array(x[0]).shape[1],np.array(x[0]).shape[1], 3)
print(input_shape)(224, 224, 3)
def visualize_img(image_batch, label_batch):
# The image is visualised from the data set.
# Parameters
# image_batch: Holds the martrix information of the images.
# label_batch: List holding the class information of the image.
plt.figure(figsize=(12, 12))
for n in range(12):
ax = plt.subplot(6,6,n+1)
plt.imshow(image_batch[n])
plt.title(np.array(list(waste_labels.keys()))[to_categorical(labels, num_classes=6)[n]==1][0].title())
plt.axis('off')visualize_img(x, labels)
# We define a generator for the Train dataset.
train = ImageDataGenerator(horizontal_flip=True,
vertical_flip=True,
validation_split=0.1,
rescale=1./255,
shear_range = 0.1,
zoom_range = 0.1,
width_shift_range= 0.1,
height_shift_range = 0.1)
# We define a generator for the test data set.
test = ImageDataGenerator(rescale=1/255,
validation_split=0.1)train_generator=train.flow_from_directory(directory=dir_path,
target_size=(target_size),
class_mode='categorical',
subset='training')
test_generator=test.flow_from_directory(directory=dir_path,
target_size=(target_size),
batch_size=251,
class_mode='categorical',
subset='validation')Found 2276 images belonging to 6 classes.
Found 251 images belonging to 6 classes.
- Sequential
- Convolution Layer (Convolotion Layer, Conv2D)
- Pooling Layer
- Activation Function Layer (Activation Layer)
- Flattening Layer
- Dense Layer
- Dropout Layer
model=Sequential()
model.add(Conv2D(filters=32, kernel_size=(3,3), padding='same', input_shape=(input_shape), activation='relu'))
model.add(MaxPooling2D(pool_size=2, strides=(2,2)))
model.add(Conv2D(filters=64, kernel_size=(3,3), padding='same', input_shape=(input_shape), activation='relu'))
model.add(MaxPooling2D(pool_size=2, strides=(2,2)))
model.add(Conv2D(filters=32, kernel_size=(3,3), padding='same', input_shape=(input_shape), activation='relu'))
model.add(MaxPooling2D(pool_size=2, strides=(2,2)))
model.add(Flatten())
model.add(Dense(units=64, activation='relu'))
model.add(Dropout(rate=0.2))
model.add(Dense(units=32, activation='relu'))
model.add(Dropout(rate=0.2))
model.add(Dense(units=6, activation='softmax'))model.summary()Model: "sequential"
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━┓ ┃ Layer (type) ┃ Output Shape ┃ Param # ┃ ┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━┩ │ conv2d (Conv2D) │ (None, 224, 224, 32) │ 896 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ max_pooling2d (MaxPooling2D) │ (None, 112, 112, 32) │ 0 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ conv2d_1 (Conv2D) │ (None, 112, 112, 64) │ 18,496 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ max_pooling2d_1 (MaxPooling2D) │ (None, 56, 56, 64) │ 0 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ conv2d_2 (Conv2D) │ (None, 56, 56, 32) │ 18,464 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ max_pooling2d_2 (MaxPooling2D) │ (None, 28, 28, 32) │ 0 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ flatten (Flatten) │ (None, 25088) │ 0 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ dense (Dense) │ (None, 64) │ 1,605,696 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ dropout (Dropout) │ (None, 64) │ 0 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ dense_1 (Dense) │ (None, 32) │ 2,080 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ dropout_1 (Dropout) │ (None, 32) │ 0 │ ├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤ │ dense_2 (Dense) │ (None, 6) │ 198 │ └──────────────────────────────────────┴─────────────────────────────┴─────────────────┘
Total params: 1,645,830 (6.28 MB)
Trainable params: 1,645,830 (6.28 MB)
Non-trainable params: 0 (0.00 B)
model.compile(loss='categorical_crossentropy',
optimizer='adam',
metrics=[tf.keras.metrics.Precision(), tf.keras.metrics.Recall(), "acc"])callbacks = [EarlyStopping(monitor='val_loss', patience=50, verbose=1, mode="min"),
ModelCheckpoint(filepath='mymodel.keras', monitor='val_loss', mode='min', save_best_only=True, save_weights_only=False, verbose=1)]history = model.fit(train_generator,
epochs=30,
validation_data=test_generator,
callbacks=callbacks,
steps_per_epoch=2276//32,
validation_steps=251//32)Epoch 1/30
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m0s�[0m 481ms/step - acc: 0.2285 - loss: 1.7641 - precision: 0.2808 - recall: 0.0068
Epoch 1: val_loss improved from inf to 1.63878, saving model to mymodel.keras
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m66s�[0m 616ms/step - acc: 0.2286 - loss: 1.7637 - precision: 0.2825 - recall: 0.0068 - val_acc: 0.2988 - val_loss: 1.6388 - val_precision: 0.8571 - val_recall: 0.0239
Epoch 2/30
�[1m 1/71�[0m �[37m━━━━━━━━━━━━━━━━━━━━�[0m �[1m2s�[0m 36ms/step - acc: 0.3125 - loss: 1.5804 - precision: 0.5000 - recall: 0.0312
Epoch 2: val_loss did not improve from 1.63878
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m1s�[0m 18ms/step - acc: 0.3125 - loss: 1.5804 - precision: 0.5000 - recall: 0.0312 - val_acc: 0.2430 - val_loss: 1.6896 - val_precision: 0.9286 - val_recall: 0.0518
Epoch 3/30
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m0s�[0m 453ms/step - acc: 0.3150 - loss: 1.6333 - precision: 0.4312 - recall: 0.0439
Epoch 3: val_loss improved from 1.63878 to 1.56500, saving model to mymodel.keras
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m57s�[0m 473ms/step - acc: 0.3154 - loss: 1.6325 - precision: 0.4338 - recall: 0.0445 - val_acc: 0.3426 - val_loss: 1.5650 - val_precision: 0.5294 - val_recall: 0.1076
Epoch 4/30
�[1m 1/71�[0m �[37m━━━━━━━━━━━━━━━━━━━━�[0m �[1m2s�[0m 39ms/step - acc: 0.3750 - loss: 1.6246 - precision: 0.5000 - recall: 0.1250
Epoch 4: val_loss improved from 1.56500 to 1.53631, saving model to mymodel.keras
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m1s�[0m 20ms/step - acc: 0.3750 - loss: 1.6246 - precision: 0.5000 - recall: 0.1250 - val_acc: 0.3426 - val_loss: 1.5363 - val_precision: 0.6000 - val_recall: 0.1076
Epoch 5/30
�[1m70/71�[0m �[32m━━━━━━━━━━━━━━━━━━━�[0m�[37m━�[0m �[1m0s�[0m 466ms/step - acc: 0.3863 - loss: 1.4515 - precision: 0.6554 - recall: 0.1565
Epoch 5: val_loss improved from 1.53631 to 1.34416, saving model to mymodel.keras
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m38s�[0m 479ms/step - acc: 0.3872 - loss: 1.4508 - precision: 0.6557 - recall: 0.1569 - val_acc: 0.4900 - val_loss: 1.3442 - val_precision: 0.7541 - val_recall: 0.1833
Epoch 6/30
�[1m 1/71�[0m �[37m━━━━━━━━━━━━━━━━━━━━�[0m �[1m2s�[0m 38ms/step - acc: 0.4375 - loss: 1.4724 - precision: 0.5833 - recall: 0.2188
Epoch 6: val_loss did not improve from 1.34416
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m2s�[0m 29ms/step - acc: 0.4375 - loss: 1.4724 - precision: 0.5833 - recall: 0.2188 - val_acc: 0.4701 - val_loss: 1.3638 - val_precision: 0.7667 - val_recall: 0.1833
Epoch 7/30
�[1m70/71�[0m �[32m━━━━━━━━━━━━━━━━━━━�[0m�[37m━�[0m �[1m0s�[0m 502ms/step - acc: 0.4406 - loss: 1.3864 - precision: 0.6929 - recall: 0.1908
Epoch 7: val_loss improved from 1.34416 to 1.27833, saving model to mymodel.keras
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m40s�[0m 518ms/step - acc: 0.4406 - loss: 1.3861 - precision: 0.6926 - recall: 0.1907 - val_acc: 0.4900 - val_loss: 1.2783 - val_precision: 0.7763 - val_recall: 0.2351
Epoch 8/30
�[1m 1/71�[0m �[37m━━━━━━━━━━━━━━━━━━━━�[0m �[1m2s�[0m 37ms/step - acc: 0.4688 - loss: 1.2193 - precision: 0.8182 - recall: 0.2812
Epoch 8: val_loss improved from 1.27833 to 1.27312, saving model to mymodel.keras
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m1s�[0m 21ms/step - acc: 0.4688 - loss: 1.2193 - precision: 0.8182 - recall: 0.2812 - val_acc: 0.4900 - val_loss: 1.2731 - val_precision: 0.7284 - val_recall: 0.2351
Epoch 9/30
�[1m70/71�[0m �[32m━━━━━━━━━━━━━━━━━━━�[0m�[37m━�[0m �[1m0s�[0m 451ms/step - acc: 0.4758 - loss: 1.3400 - precision: 0.6876 - recall: 0.2318
Epoch 9: val_loss improved from 1.27312 to 1.25208, saving model to mymodel.keras
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m36s�[0m 464ms/step - acc: 0.4755 - loss: 1.3402 - precision: 0.6875 - recall: 0.2311 - val_acc: 0.4741 - val_loss: 1.2521 - val_precision: 0.7831 - val_recall: 0.2590
Epoch 10/30
�[1m 1/71�[0m �[37m━━━━━━━━━━━━━━━━━━━━�[0m �[1m2s�[0m 36ms/step - acc: 0.5625 - loss: 1.3933 - precision: 0.8571 - recall: 0.3750
Epoch 10: val_loss improved from 1.25208 to 1.23198, saving model to mymodel.keras
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m3s�[0m 39ms/step - acc: 0.5625 - loss: 1.3933 - precision: 0.8571 - recall: 0.3750 - val_acc: 0.5139 - val_loss: 1.2320 - val_precision: 0.7619 - val_recall: 0.2550
Epoch 11/30
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m0s�[0m 449ms/step - acc: 0.4885 - loss: 1.2907 - precision: 0.7181 - recall: 0.2690
Epoch 11: val_loss did not improve from 1.23198
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m37s�[0m 468ms/step - acc: 0.4887 - loss: 1.2906 - precision: 0.7182 - recall: 0.2691 - val_acc: 0.5179 - val_loss: 1.2893 - val_precision: 0.6667 - val_recall: 0.2311
Epoch 12/30
�[1m 1/71�[0m �[37m━━━━━━━━━━━━━━━━━━━━�[0m �[1m2s�[0m 37ms/step - acc: 0.6875 - loss: 0.9264 - precision: 0.8462 - recall: 0.3438
Epoch 12: val_loss did not improve from 1.23198
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m1s�[0m 18ms/step - acc: 0.6875 - loss: 0.9264 - precision: 0.8462 - recall: 0.3438 - val_acc: 0.5339 - val_loss: 1.2536 - val_precision: 0.7308 - val_recall: 0.2271
Epoch 13/30
�[1m70/71�[0m �[32m━━━━━━━━━━━━━━━━━━━�[0m�[37m━�[0m �[1m0s�[0m 459ms/step - acc: 0.5104 - loss: 1.2583 - precision: 0.7330 - recall: 0.2811
Epoch 13: val_loss did not improve from 1.23198
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m40s�[0m 471ms/step - acc: 0.5098 - loss: 1.2592 - precision: 0.7322 - recall: 0.2807 - val_acc: 0.4940 - val_loss: 1.2446 - val_precision: 0.7595 - val_recall: 0.2390
Epoch 14/30
�[1m 1/71�[0m �[37m━━━━━━━━━━━━━━━━━━━━�[0m �[1m3s�[0m 43ms/step - acc: 0.5625 - loss: 1.2009 - precision: 0.8889 - recall: 0.2500
Epoch 14: val_loss did not improve from 1.23198
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m1s�[0m 19ms/step - acc: 0.5625 - loss: 1.2009 - precision: 0.8889 - recall: 0.2500 - val_acc: 0.5100 - val_loss: 1.2416 - val_precision: 0.7561 - val_recall: 0.2470
Epoch 15/30
�[1m70/71�[0m �[32m━━━━━━━━━━━━━━━━━━━�[0m�[37m━�[0m �[1m0s�[0m 456ms/step - acc: 0.5124 - loss: 1.2504 - precision: 0.7321 - recall: 0.2933
Epoch 15: val_loss did not improve from 1.23198
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m40s�[0m 468ms/step - acc: 0.5120 - loss: 1.2514 - precision: 0.7314 - recall: 0.2930 - val_acc: 0.5139 - val_loss: 1.2967 - val_precision: 0.8033 - val_recall: 0.1952
Epoch 16/30
�[1m 1/71�[0m �[37m━━━━━━━━━━━━━━━━━━━━�[0m �[1m2s�[0m 36ms/step - acc: 0.4062 - loss: 1.3972 - precision: 0.6000 - recall: 0.0938
Epoch 16: val_loss did not improve from 1.23198
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m1s�[0m 18ms/step - acc: 0.4062 - loss: 1.3972 - precision: 0.6000 - recall: 0.0938 - val_acc: 0.5060 - val_loss: 1.2945 - val_precision: 0.8448 - val_recall: 0.1952
Epoch 17/30
�[1m70/71�[0m �[32m━━━━━━━━━━━━━━━━━━━�[0m�[37m━�[0m �[1m0s�[0m 501ms/step - acc: 0.5299 - loss: 1.2156 - precision: 0.7394 - recall: 0.2892
Epoch 17: val_loss improved from 1.23198 to 1.20952, saving model to mymodel.keras
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m41s�[0m 514ms/step - acc: 0.5296 - loss: 1.2160 - precision: 0.7393 - recall: 0.2898 - val_acc: 0.5060 - val_loss: 1.2095 - val_precision: 0.7935 - val_recall: 0.2908
Epoch 18/30
�[1m 1/71�[0m �[37m━━━━━━━━━━━━━━━━━━━━�[0m �[1m2s�[0m 37ms/step - acc: 0.6875 - loss: 1.0896 - precision: 0.9333 - recall: 0.4375
Epoch 18: val_loss did not improve from 1.20952
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m1s�[0m 19ms/step - acc: 0.6875 - loss: 1.0896 - precision: 0.9333 - recall: 0.4375 - val_acc: 0.5060 - val_loss: 1.2255 - val_precision: 0.7976 - val_recall: 0.2669
Epoch 19/30
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m0s�[0m 451ms/step - acc: 0.5461 - loss: 1.1814 - precision: 0.7829 - recall: 0.3232
Epoch 19: val_loss improved from 1.20952 to 1.18194, saving model to mymodel.keras
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m38s�[0m 490ms/step - acc: 0.5460 - loss: 1.1816 - precision: 0.7826 - recall: 0.3233 - val_acc: 0.5378 - val_loss: 1.1819 - val_precision: 0.7130 - val_recall: 0.3267
Epoch 20/30
�[1m 1/71�[0m �[37m━━━━━━━━━━━━━━━━━━━━�[0m �[1m3s�[0m 46ms/step - acc: 0.5000 - loss: 1.3044 - precision: 0.6429 - recall: 0.2812
Epoch 20: val_loss did not improve from 1.18194
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m1s�[0m 20ms/step - acc: 0.5000 - loss: 1.3044 - precision: 0.6429 - recall: 0.2812 - val_acc: 0.5458 - val_loss: 1.1871 - val_precision: 0.7522 - val_recall: 0.3386
Epoch 21/30
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m0s�[0m 456ms/step - acc: 0.5233 - loss: 1.2451 - precision: 0.7361 - recall: 0.3123
Epoch 21: val_loss improved from 1.18194 to 1.16648, saving model to mymodel.keras
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m39s�[0m 479ms/step - acc: 0.5234 - loss: 1.2446 - precision: 0.7362 - recall: 0.3125 - val_acc: 0.5339 - val_loss: 1.1665 - val_precision: 0.7273 - val_recall: 0.3506
Epoch 22/30
�[1m 1/71�[0m �[37m━━━━━━━━━━━━━━━━━━━━�[0m �[1m2s�[0m 41ms/step - acc: 0.6875 - loss: 0.9779 - precision: 0.7692 - recall: 0.3125
Epoch 22: val_loss did not improve from 1.16648
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m2s�[0m 23ms/step - acc: 0.6875 - loss: 0.9779 - precision: 0.7692 - recall: 0.3125 - val_acc: 0.5139 - val_loss: 1.1673 - val_precision: 0.7345 - val_recall: 0.3307
Epoch 23/30
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m0s�[0m 452ms/step - acc: 0.5492 - loss: 1.1750 - precision: 0.7495 - recall: 0.3571
Epoch 23: val_loss did not improve from 1.16648
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m37s�[0m 470ms/step - acc: 0.5492 - loss: 1.1749 - precision: 0.7494 - recall: 0.3570 - val_acc: 0.5378 - val_loss: 1.2278 - val_precision: 0.7212 - val_recall: 0.2988
Epoch 24/30
�[1m 1/71�[0m �[37m━━━━━━━━━━━━━━━━━━━━�[0m �[1m2s�[0m 38ms/step - acc: 0.6562 - loss: 1.1657 - precision: 0.7333 - recall: 0.3438
Epoch 24: val_loss did not improve from 1.16648
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m1s�[0m 19ms/step - acc: 0.6562 - loss: 1.1657 - precision: 0.7333 - recall: 0.3438 - val_acc: 0.5139 - val_loss: 1.2666 - val_precision: 0.7500 - val_recall: 0.2869
Epoch 25/30
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m0s�[0m 444ms/step - acc: 0.5402 - loss: 1.1581 - precision: 0.7843 - recall: 0.3268
Epoch 25: val_loss did not improve from 1.16648
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m40s�[0m 462ms/step - acc: 0.5403 - loss: 1.1580 - precision: 0.7841 - recall: 0.3272 - val_acc: 0.5299 - val_loss: 1.2099 - val_precision: 0.6940 - val_recall: 0.3705
Epoch 26/30
�[1m 1/71�[0m �[37m━━━━━━━━━━━━━━━━━━━━�[0m �[1m2s�[0m 38ms/step - acc: 0.5938 - loss: 1.0940 - precision: 0.7647 - recall: 0.4062
Epoch 26: val_loss did not improve from 1.16648
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m1s�[0m 18ms/step - acc: 0.5938 - loss: 1.0940 - precision: 0.7647 - recall: 0.4062 - val_acc: 0.5060 - val_loss: 1.3021 - val_precision: 0.6408 - val_recall: 0.3625
Epoch 27/30
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m0s�[0m 456ms/step - acc: 0.5561 - loss: 1.1583 - precision: 0.7306 - recall: 0.3685
Epoch 27: val_loss improved from 1.16648 to 1.15336, saving model to mymodel.keras
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m41s�[0m 494ms/step - acc: 0.5563 - loss: 1.1579 - precision: 0.7307 - recall: 0.3686 - val_acc: 0.5498 - val_loss: 1.1534 - val_precision: 0.6940 - val_recall: 0.3705
Epoch 28/30
�[1m 1/71�[0m �[37m━━━━━━━━━━━━━━━━━━━━�[0m �[1m3s�[0m 52ms/step - acc: 0.5938 - loss: 1.2291 - precision: 0.8235 - recall: 0.4375
Epoch 28: val_loss did not improve from 1.15336
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m1s�[0m 18ms/step - acc: 0.5938 - loss: 1.2291 - precision: 0.8235 - recall: 0.4375 - val_acc: 0.5378 - val_loss: 1.1654 - val_precision: 0.6906 - val_recall: 0.3825
Epoch 29/30
�[1m70/71�[0m �[32m━━━━━━━━━━━━━━━━━━━�[0m�[37m━�[0m �[1m0s�[0m 476ms/step - acc: 0.5811 - loss: 1.0788 - precision: 0.7752 - recall: 0.3975
Epoch 29: val_loss did not improve from 1.15336
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m40s�[0m 488ms/step - acc: 0.5813 - loss: 1.0793 - precision: 0.7751 - recall: 0.3975 - val_acc: 0.5458 - val_loss: 1.1659 - val_precision: 0.7607 - val_recall: 0.3546
Epoch 30/30
�[1m 1/71�[0m �[37m━━━━━━━━━━━━━━━━━━━━�[0m �[1m2s�[0m 36ms/step - acc: 0.6875 - loss: 0.9615 - precision: 0.8500 - recall: 0.5312
Epoch 30: val_loss did not improve from 1.15336
�[1m71/71�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m1s�[0m 18ms/step - acc: 0.6875 - loss: 0.9615 - precision: 0.8500 - recall: 0.5312 - val_acc: 0.5339 - val_loss: 1.1806 - val_precision: 0.7165 - val_recall: 0.3625
# Accuracy Graph
plt.figure(figsize=(20, 5))
plt.subplot(1, 2, 1)
plt.plot(history.history['acc'], label ='Training Accuracy')
plt.plot(history.history['val_acc'], label ='Validation Accuracy')
plt.legend(loc='lower right')
plt.xlabel('Epoch', fontsize=16)
plt.ylabel('Accuracy', fontsize=16)
plt.ylim([min(plt.ylim()),1])
plt.title('Training and Validation Accuracy', fontsize=16)
# Loss Graph
plt.subplot(1, 2, 2)
plt.plot(history.history['loss'], label ='Training Loss')
plt.plot(history.history['val_loss'], label ='Validation Loss')
plt.legend(loc='upper right')
plt.xlabel('Epoch', fontsize=16)
plt.ylabel('Loss', fontsize=16)
plt.ylim([0, max(plt.ylim())])
plt.title('Training and Validation Loss', fontsize=16)
plt.show()
loss, precision, recall, acc = model.evaluate(test_generator, batch_size=32)
print("\nTest Accuracy: %.1f%%" % (100.0 * acc))
print("\nTest Loss: %.1f%%" % (100.0 * loss))
print("\nTest Precision: %.1f%%" % (100.0 * precision))
print("\nTest Recall: %.1f%%" % (100.0 * recall))�[1m1/1�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m1s�[0m 1s/step - acc: 0.5339 - loss: 1.1806 - precision: 0.7165 - recall: 0.3625
Test Accuracy: 53.4%
Test Loss: 118.1%
Test Precision: 71.7%
Test Recall: 36.3%
# Classification Report
x_test, y_test = next(test_generator)
y_pred = model.predict(x_test)
y_pred = np.argmax(y_pred, axis=1)
y_test = np.argmax(y_test, axis=1)
y_pred�[1m8/8�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m2s�[0m 169ms/step
array([3, 4, 0, 2, 4, 3, 1, 2, 0, 0, 0, 3, 3, 2, 3, 3, 1, 2, 3, 2, 3, 3,
0, 3, 1, 0, 2, 1, 3, 4, 1, 2, 1, 3, 3, 2, 3, 2, 2, 1, 2, 1, 1, 1,
3, 2, 1, 3, 1, 3, 3, 1, 0, 3, 3, 4, 2, 1, 2, 4, 1, 2, 1, 3, 1, 3,
2, 1, 3, 1, 1, 1, 2, 1, 1, 1, 1, 1, 1, 4, 4, 2, 4, 3, 0, 1, 2, 2,
1, 2, 2, 3, 2, 3, 2, 3, 0, 1, 2, 2, 3, 2, 2, 3, 0, 3, 1, 1, 4, 3,
3, 4, 1, 3, 3, 1, 1, 0, 2, 2, 3, 2, 4, 3, 3, 1, 0, 4, 1, 3, 0, 3,
4, 1, 0, 4, 2, 0, 0, 2, 0, 2, 0, 3, 3, 3, 0, 0, 4, 3, 3, 0, 2, 3,
1, 3, 1, 0, 1, 2, 1, 3, 1, 0, 4, 3, 3, 0, 3, 0, 3, 0, 1, 3, 2, 0,
1, 2, 4, 4, 0, 3, 1, 2, 3, 3, 0, 3, 3, 1, 2, 2, 3, 5, 3, 2, 0, 2,
1, 2, 4, 3, 0, 1, 3, 2, 1, 3, 0, 1, 0, 1, 1, 2, 2, 3, 3, 1, 3, 2,
2, 1, 0, 1, 2, 1, 0, 3, 3, 3, 2, 3, 2, 2, 2, 2, 1, 1, 4, 2, 2, 0,
4, 2, 3, 1, 2, 3, 3, 2, 2])
target_names = list(waste_labels.keys())print(classification_report(y_test, y_pred, target_names=target_names)) precision recall f1-score support
cardboard 0.64 0.57 0.61 40
glass 0.40 0.48 0.44 50
metal 0.48 0.71 0.57 41
paper 0.62 0.76 0.69 59
plastic 0.62 0.27 0.38 48
trash 0.00 0.00 0.00 13
accuracy 0.53 251
macro avg 0.46 0.47 0.45 251
weighted avg 0.52 0.53 0.51 251
# Confusion Matrix
cm = confusion_matrix(y_test, y_pred)
def plot_confusion_matrix(cm, classes,
normalize=False,
title='Confusion Matrix',
cmap=plt.cm.Blues):
# This function prints and plots the confusion matrix.
# Normalization can be applied by setting `normalize=True`.
if normalize:
cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]
plt.figure(figsize=(8,6))
plt.imshow(cm, interpolation='nearest', cmap=cmap)
plt.title(title)
plt.colorbar()
tick_marks = np.arange(len(classes))
plt.xticks(tick_marks, classes, rotation=45)
plt.yticks(tick_marks, classes)
fmt = '.2f' if normalize else 'd'
thresh = cm.max() / 2.
for i, j in itertools.product(range(cm.shape[0]), range(cm.shape[1])):
plt.text(j, i, format(cm[i, j], fmt),
horizontalalignment="center",
color="white" if cm[i, j] > thresh else "black")
plt.tight_layout()
plt.ylabel('True Label', fontweight="bold")
plt.xlabel('Predict Label', fontweight="bold")plot_confusion_matrix(cm, waste_labels.keys(),
title='Confusion Matrix',
cmap=plt.cm.OrRd)
waste_labels = {0: 'cardboard', 1: 'glass', 2: 'metal', 3: 'paper', 4: 'plastic', 5: 'trash'}def model_testing(path):
# By reading the image in the directory where the images are located, the model is used to predict which class it belongs to.
# Parameters
# path: The directory where the images are located.
# Return:
# img: Image
# p: Prediction probabilities
# predicted_class: Predicted class
img = image.load_img(path, target_size=(target_size))
img = image.img_to_array(img, dtype=np.uint8)
img=np.array(img)/255.0
p=model.predict(img.reshape(1, 224, 224, 3))
predicted_class = np.argmax(p[0])
return img, p, predicted_class# Test images
img1, p1, predicted_class1 = model_testing('/content/drive/MyDrive/Garbage-Classification/metal/metal1.jpg')
img2, p2, predicted_class2 = model_testing('/content/drive/MyDrive/Garbage-Classification/glass/glass5.jpg')
img3, p3, predicted_class3 = model_testing('/content/drive/MyDrive/Garbage-Classification/cardboard/cardboard13.jpg')
img4, p4, predicted_class4 = model_testing('/content/drive/MyDrive/Garbage-Classification/paper/paper106.jpg')
img5, p5, predicted_class5 = model_testing('/content/drive/MyDrive/Garbage-Classification/plastic/plastic14.jpg')
img6, p6, predicted_class6 = model_testing('/content/drive/MyDrive/Garbage-Classification/trash/trash26.jpg')
# Displaying test images
plt.figure(figsize=(12, 8))
plt.subplot(2, 3, 1)
plt.axis('off')
plt.imshow(img1.squeeze())
plt.title(f"Max Prob: {np.max(p1[0])}\nClass: {waste_labels[predicted_class1]}")
plt.subplot(2, 3, 2)
plt.axis('off')
plt.imshow(img2.squeeze())
plt.title(f"Max Prob: {np.max(p2[0])}\nClass: {waste_labels[predicted_class2]}")
plt.subplot(2, 3, 3)
plt.axis('off')
plt.imshow(img3.squeeze())
plt.title(f"Max Prob: {np.max(p3[0])}\nClass: {waste_labels[predicted_class3]}")
plt.subplot(2, 3, 4)
plt.axis('off')
plt.imshow(img4.squeeze())
plt.title(f"Max Prob: {np.max(p4[0])}\nClass: {waste_labels[predicted_class4]}")
plt.subplot(2, 3, 5)
plt.axis('off')
plt.imshow(img5.squeeze())
plt.title(f"Max Prob: {np.max(p5[0])}\nClass: {waste_labels[predicted_class5]}")
plt.subplot(2, 3, 6)
plt.axis('off')
plt.imshow(img6.squeeze())
plt.title(f"Max Prob: {np.max(p6[0])}\nClass: {waste_labels[predicted_class6]}")
plt.show()�[1m1/1�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m1s�[0m 617ms/step
�[1m1/1�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m0s�[0m 25ms/step
�[1m1/1�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m0s�[0m 22ms/step
�[1m1/1�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m0s�[0m 28ms/step
�[1m1/1�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m0s�[0m 26ms/step
�[1m1/1�[0m �[32m━━━━━━━━━━━━━━━━━━━━�[0m�[37m�[0m �[1m0s�[0m 26ms/step
