In [0]:
import numpy as np
import matplotlib.pyplot as plt
from keras.preprocessing.image import ImageDataGenerator, load_img, img_to_array, array_to_img
from keras.layers import Conv2D, Flatten, MaxPooling2D, Dense
from keras.models import Sequential
import tensorflow as tf
import glob, os, random
In [0]:
from google.colab import drive
drive.mount('/content/drive')
Mounted at /content/drive
In [0]:
os.getcwd()
Out[0]:
'/content'
In [0]:
base_path = 'drive/My Drive/Colab Notebooks/Dataset'
img_list = glob.glob(os.path.join(base_path, '*/*.jpg'))
print(len(img_list))
2538
In [0]:
for i, img_path in enumerate(random.sample(img_list, 6)):
img = load_img(img_path)
img = img_to_array(img, dtype=np.uint8)
plt.subplot(2, 3, i+1)
plt.imshow(img.squeeze())
In [0]:
train_datagen = ImageDataGenerator(
rescale=1./255,
shear_range=0.1,
zoom_range=0.1,
width_shift_range=0.1,
height_shift_range=0.1,
horizontal_flip=True,
vertical_flip=True,
validation_split=0.1
)
test_datagen = ImageDataGenerator(
rescale=1./255,
validation_split=0.1
)
train_generator = train_datagen.flow_from_directory(
base_path,
target_size=(224, 224),
batch_size=16,
class_mode='categorical',
subset='training',
seed=0
)
validation_generator = test_datagen.flow_from_directory(
base_path,
target_size=(224, 224),
batch_size=16,
class_mode='categorical',
subset='validation',
seed=0
)
labels = (train_generator.class_indices)
labels = dict((v,k) for k,v in labels.items())
print(labels)
Found 2286 images belonging to 6 classes.
Found 252 images belonging to 6 classes.
{0: 'cardboard', 1: 'glass', 2: 'metal', 3: 'paper', 4: 'plastic', 5: 'trash'}
In [0]:
IMG_SHAPE = (224,224,3)
base_model = tf.keras.applications.MobileNetV2(input_shape=IMG_SHAPE,
include_top=False,
weights='imagenet')
WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/resource_variable_ops.py:435: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version. Instructions for updating: Colocations handled automatically by placer. Downloading data from https://github.com/JonathanCMitchell/mobilenet_v2_keras/releases/download/v1.1/mobilenet_v2_weights_tf_dim_ordering_tf_kernels_1.0_224_no_top.h5 9412608/9406464 [==============================] - 1s 0us/step
In [0]:
model = Sequential([
Conv2D(filters=32, kernel_size=3, padding='same', activation='relu', input_shape=(224, 224, 3)),
MaxPooling2D(pool_size=2),
Conv2D(filters=64, kernel_size=3, padding='same', activation='relu'),
MaxPooling2D(pool_size=2),
Conv2D(filters=32, kernel_size=3, padding='same', activation='relu'),
MaxPooling2D(pool_size=2),
Conv2D(filters=32, kernel_size=3, padding='same', activation='relu'),
MaxPooling2D(pool_size=2),
Flatten(),
Dense(64, activation='relu'),
Dense(6, activation='softmax')
])
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['acc'])
model.summary()
_________________________________________________________________ Layer (type) Output Shape Param # ================================================================= conv2d_9 (Conv2D) (None, 224, 224, 32) 896 _________________________________________________________________ max_pooling2d_9 (MaxPooling2 (None, 112, 112, 32) 0 _________________________________________________________________ conv2d_10 (Conv2D) (None, 112, 112, 64) 18496 _________________________________________________________________ max_pooling2d_10 (MaxPooling (None, 56, 56, 64) 0 _________________________________________________________________ conv2d_11 (Conv2D) (None, 56, 56, 32) 18464 _________________________________________________________________ max_pooling2d_11 (MaxPooling (None, 28, 28, 32) 0 _________________________________________________________________ conv2d_12 (Conv2D) (None, 28, 28, 32) 9248 _________________________________________________________________ max_pooling2d_12 (MaxPooling (None, 14, 14, 32) 0 _________________________________________________________________ flatten_3 (Flatten) (None, 6272) 0 _________________________________________________________________ dense_5 (Dense) (None, 64) 401472 _________________________________________________________________ dense_6 (Dense) (None, 6) 390 ================================================================= Total params: 448,966 Trainable params: 448,966 Non-trainable params: 0 _________________________________________________________________
In [0]:
batch_size = 32
epochs = 20
history = model.fit_generator(train_generator,
steps_per_epoch=train_generator.n // train_generator.batch_size,
epochs = epochs,
workers = 4,
validation_data=validation_generator,
validation_steps=validation_generator.n // validation_generator.batch_size,
verbose=1)
WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version. Instructions for updating: Use tf.cast instead. Epoch 1/20 142/142 [==============================] - 285s 2s/step - loss: 1.6548 - acc: 0.2764 - val_loss: 1.5253 - val_acc: 0.3583 Epoch 2/20 142/142 [==============================] - 266s 2s/step - loss: 1.4497 - acc: 0.4128 - val_loss: 1.4759 - val_acc: 0.4322 Epoch 3/20 142/142 [==============================] - 265s 2s/step - loss: 1.3832 - acc: 0.4317 - val_loss: 1.3475 - val_acc: 0.4153 Epoch 4/20 142/142 [==============================] - 273s 2s/step - loss: 1.3177 - acc: 0.4608 - val_loss: 1.3914 - val_acc: 0.4449 Epoch 5/20 142/142 [==============================] - 274s 2s/step - loss: 1.2473 - acc: 0.4993 - val_loss: 1.6125 - val_acc: 0.4449 Epoch 6/20 142/142 [==============================] - 274s 2s/step - loss: 1.2191 - acc: 0.5155 - val_loss: 1.3080 - val_acc: 0.4746 Epoch 7/20 142/142 [==============================] - 283s 2s/step - loss: 1.1079 - acc: 0.5741 - val_loss: 1.2368 - val_acc: 0.5381 Epoch 8/20 142/142 [==============================] - 292s 2s/step - loss: 1.0701 - acc: 0.5970 - val_loss: 1.1457 - val_acc: 0.5466 Epoch 9/20 142/142 [==============================] - 294s 2s/step - loss: 1.0157 - acc: 0.6190 - val_loss: 1.2068 - val_acc: 0.5466 Epoch 10/20 142/142 [==============================] - 294s 2s/step - loss: 1.0242 - acc: 0.6095 - val_loss: 1.0715 - val_acc: 0.5932 Epoch 11/20 142/142 [==============================] - 290s 2s/step - loss: 0.9655 - acc: 0.6439 - val_loss: 1.1762 - val_acc: 0.4958 Epoch 12/20 142/142 [==============================] - 290s 2s/step - loss: 0.9477 - acc: 0.6395 - val_loss: 1.1370 - val_acc: 0.5847 Epoch 13/20 142/142 [==============================] - 291s 2s/step - loss: 0.9054 - acc: 0.6605 - val_loss: 0.9937 - val_acc: 0.6441 Epoch 14/20 142/142 [==============================] - 293s 2s/step - loss: 0.8833 - acc: 0.6701 - val_loss: 0.9857 - val_acc: 0.6398 Epoch 15/20 142/142 [==============================] - 295s 2s/step - loss: 0.8402 - acc: 0.6845 - val_loss: 0.9360 - val_acc: 0.6525 Epoch 16/20 142/142 [==============================] - 296s 2s/step - loss: 0.7595 - acc: 0.7176 - val_loss: 0.8945 - val_acc: 0.6822 Epoch 17/20 142/142 [==============================] - 290s 2s/step - loss: 0.7773 - acc: 0.7020 - val_loss: 1.1326 - val_acc: 0.6167 Epoch 18/20 142/142 [==============================] - 290s 2s/step - loss: 0.7919 - acc: 0.7102 - val_loss: 1.0370 - val_acc: 0.6568 Epoch 19/20 142/142 [==============================] - 288s 2s/step - loss: 0.7329 - acc: 0.7317 - val_loss: 1.0717 - val_acc: 0.6186 Epoch 20/20 142/142 [==============================] - 288s 2s/step - loss: 0.7194 - acc: 0.7469 - val_loss: 0.9630 - val_acc: 0.6398
In [0]:
batch_size = 32
epochs = 20
history = model.fit_generator(train_generator,
steps_per_epoch=train_generator.n // train_generator.batch_size,
epochs = epochs,
workers = 4,
validation_data=validation_generator,
validation_steps=validation_generator.n // validation_generator.batch_size,
verbose=1)
# list all data in history
print(history.history.keys())
# summarize history for accuracy
plt.plot(history.history['acc'])
plt.plot(history.history['val_acc'])
plt.title('model accuracy')
plt.ylabel('accuracy')
plt.xlabel('epoch')
plt.legend(['train', 'test'], loc='upper left')
plt.show()
# summarize history for loss
plt.plot(history.history['loss'])
plt.plot(history.history['val_loss'])
plt.title('model loss')
plt.ylabel('loss')
plt.xlabel('epoch')
plt.legend(['train', 'test'], loc='upper left')
plt.show()
Epoch 1/20 142/142 [==============================] - 303s 2s/step - loss: 0.6759 - acc: 0.7458 - val_loss: 0.9028 - val_acc: 0.6695 Epoch 2/20 142/142 [==============================] - 296s 2s/step - loss: 0.7029 - acc: 0.7481 - val_loss: 0.9696 - val_acc: 0.6568 Epoch 3/20 142/142 [==============================] - 296s 2s/step - loss: 0.6798 - acc: 0.7594 - val_loss: 0.7899 - val_acc: 0.7034 Epoch 4/20 142/142 [==============================] - 297s 2s/step - loss: 0.6373 - acc: 0.7671 - val_loss: 0.8871 - val_acc: 0.6864 Epoch 5/20 142/142 [==============================] - 296s 2s/step - loss: 0.6175 - acc: 0.7701 - val_loss: 0.7405 - val_acc: 0.7203 Epoch 6/20 142/142 [==============================] - 295s 2s/step - loss: 0.6084 - acc: 0.7723 - val_loss: 0.9270 - val_acc: 0.6864 Epoch 7/20 142/142 [==============================] - 294s 2s/step - loss: 0.6163 - acc: 0.7733 - val_loss: 0.8852 - val_acc: 0.6907 Epoch 8/20 142/142 [==============================] - 295s 2s/step - loss: 0.5981 - acc: 0.7899 - val_loss: 0.9267 - val_acc: 0.6833 Epoch 9/20 142/142 [==============================] - 295s 2s/step - loss: 0.5955 - acc: 0.7736 - val_loss: 0.9368 - val_acc: 0.7119 Epoch 10/20 142/142 [==============================] - 294s 2s/step - loss: 0.5667 - acc: 0.7884 - val_loss: 0.8670 - val_acc: 0.6822 Epoch 11/20 142/142 [==============================] - 286s 2s/step - loss: 0.5510 - acc: 0.7970 - val_loss: 0.7389 - val_acc: 0.7458 Epoch 12/20 142/142 [==============================] - 285s 2s/step - loss: 0.5824 - acc: 0.7885 - val_loss: 0.9318 - val_acc: 0.7203 Epoch 13/20 142/142 [==============================] - 290s 2s/step - loss: 0.5273 - acc: 0.8031 - val_loss: 0.8069 - val_acc: 0.7288 Epoch 14/20 142/142 [==============================] - 290s 2s/step - loss: 0.5445 - acc: 0.7887 - val_loss: 0.9021 - val_acc: 0.6949 Epoch 15/20 142/142 [==============================] - 291s 2s/step - loss: 0.5372 - acc: 0.8043 - val_loss: 0.8375 - val_acc: 0.7076 Epoch 16/20 142/142 [==============================] - 290s 2s/step - loss: 0.5050 - acc: 0.8101 - val_loss: 0.9467 - val_acc: 0.6780 Epoch 17/20 142/142 [==============================] - 290s 2s/step - loss: 0.5049 - acc: 0.8207 - val_loss: 0.9277 - val_acc: 0.7161 Epoch 18/20 142/142 [==============================] - 290s 2s/step - loss: 0.4607 - acc: 0.8172 - val_loss: 1.0004 - val_acc: 0.6822 Epoch 19/20 142/142 [==============================] - 291s 2s/step - loss: 0.5227 - acc: 0.8053 - val_loss: 0.9447 - val_acc: 0.6822 Epoch 20/20 142/142 [==============================] - 293s 2s/step - loss: 0.4580 - acc: 0.8353 - val_loss: 0.7916 - val_acc: 0.7754 dict_keys(['val_loss', 'val_acc', 'loss', 'acc'])
In [0]:
test_x, test_y = validation_generator.__getitem__(1)
preds = model.predict(test_x)
plt.figure(figsize=(16, 16))
for i in range(16):
plt.subplot(4, 4, i+1)
plt.title('pred:%s / truth:%s' % (labels[np.argmax(preds[i])], labels[np.argmax(test_y[i])]))
plt.imshow(test_x[i])
