In [3]:
display_image_data(config.DATASET_PATH)
Out[3]:
from drig.preprocessors import ImageToArrayPreprocessor, UniformAspectPreprocessor
from drig.datum import ImageDatumLoader
from drig.datum_config import Flowers17Config as config
from drig.config import TinyVGGNetImage, ImageCast
from sklearn.preprocessing import LabelBinarizer, LabelEncoder
from sklearn.model_selection import train_test_split
from drig.networks import TinyVGGNet
from keras.optimizers import SGD
from sklearn.metrics import classification_report
from drig.utils import plot_training_metrics, plot_network, \
display_image_data, display_prediction, display_image, \
visualize_network, plot_confusion_mesh, confusion_mesh, random_image, display_prediction, preprocess_image
from keras.preprocessing.image import ImageDataGenerator
display_image_data(config.DATASET_PATH)
config.IMAGE_PREPROCESSING_HEIGHT,config.IMAGE_PREPROCESSING_WIDTH
(64, 64)
uniform_aspect_preproc=UniformAspectPreprocessor(config.IMAGE_PREPROCESSING_HEIGHT,
config.IMAGE_PREPROCESSING_WIDTH)
image_to_array_preproc=ImageToArrayPreprocessor()
preprocessors=[uniform_aspect_preproc, image_to_array_preproc]
data_loader=ImageDatumLoader(preprocessors=preprocessors)
data, labels= data_loader.load(dataset_path=config.DATASET_PATH,
class_index=config.CLASS_INDEX,
input_cast=ImageCast.RGB_64x64)
Loading Images :100% |❆❆❆❆❆❆❆❆❆❆❆❆❆❆❆❆❆❆❆❆❆❆❆❆❆❆❆❆❆| 1360 of 1360 Time: 0:00:05 INFO:root:PROCESSED 1360/1360 IMAGES. INFO:root:MEMORY SIZE OCCUPIED: 65.28 MB
classes=data_loader.classes
train_x, test_x, train_y, test_y= train_test_split(data, labels, test_size=config.TEST_SIZE, random_state=42)
label_binarizer=LabelBinarizer()
train_y=label_binarizer.fit_transform(train_y)
test_y=label_binarizer.transform(test_y)
augmenter=ImageDataGenerator(rotation_range=30, width_shift_range=0.2, height_shift_range=0.2,
shear_range=0.2, zoom_range=0.2, horizontal_flip=True, fill_mode='nearest')
refiner=SGD(learning_rate=config.ALPHA)
net=TinyVGGNet.compose(height=TinyVGGNetImage.HEIGHT,
width=TinyVGGNetImage.WIDTH,
depth=TinyVGGNetImage.DEPTH,
classes=len(classes))
net.compile(loss="categorical_crossentropy", optimizer=refiner, metrics=["accuracy"])
visualize_network(net)
Z=net.fit(augmenter.flow(train_x, train_y, batch_size=config.BATCH_SIZE), validation_data=(test_x, test_y),
steps_per_epoch=len(train_x)//config.BATCH_SIZE, epochs=config.EPOCHS, verbose=1)
Epoch 1/100 15/15 [==============================] - 3s 160ms/step - loss: 3.8052 - accuracy: 0.1219 - val_loss: 2.8331 - val_accuracy: 0.0912 Epoch 2/100 15/15 [==============================] - 2s 137ms/step - loss: 2.6940 - accuracy: 0.2708 - val_loss: 3.0199 - val_accuracy: 0.0824 Epoch 3/100 15/15 [==============================] - 2s 141ms/step - loss: 2.3786 - accuracy: 0.3006 - val_loss: 3.1960 - val_accuracy: 0.0853 Epoch 4/100 15/15 [==============================] - 2s 138ms/step - loss: 2.0820 - accuracy: 0.3958 - val_loss: 3.3537 - val_accuracy: 0.1118 Epoch 5/100 15/15 [==============================] - 2s 137ms/step - loss: 2.0095 - accuracy: 0.4032 - val_loss: 3.4977 - val_accuracy: 0.0735 Epoch 6/100 15/15 [==============================] - 2s 136ms/step - loss: 1.9048 - accuracy: 0.4284 - val_loss: 3.5764 - val_accuracy: 0.0676 Epoch 7/100 15/15 [==============================] - 2s 136ms/step - loss: 1.8873 - accuracy: 0.4338 - val_loss: 3.6688 - val_accuracy: 0.0588 Epoch 8/100 15/15 [==============================] - 2s 135ms/step - loss: 1.7055 - accuracy: 0.4750 - val_loss: 4.0738 - val_accuracy: 0.0676 Epoch 9/100 15/15 [==============================] - 2s 137ms/step - loss: 1.7483 - accuracy: 0.4538 - val_loss: 3.8867 - val_accuracy: 0.0676 Epoch 10/100 15/15 [==============================] - 2s 135ms/step - loss: 1.6294 - accuracy: 0.4719 - val_loss: 3.8803 - val_accuracy: 0.0676 Epoch 11/100 15/15 [==============================] - 2s 136ms/step - loss: 1.5682 - accuracy: 0.4991 - val_loss: 3.9806 - val_accuracy: 0.0706 Epoch 12/100 15/15 [==============================] - 2s 135ms/step - loss: 1.3664 - accuracy: 0.5315 - val_loss: 3.5731 - val_accuracy: 0.0794 Epoch 13/100 15/15 [==============================] - 2s 134ms/step - loss: 1.4527 - accuracy: 0.4880 - val_loss: 3.7785 - val_accuracy: 0.0765 Epoch 14/100 15/15 [==============================] - 2s 135ms/step - loss: 1.3927 - accuracy: 0.5254 - val_loss: 3.6131 - val_accuracy: 0.0971 Epoch 15/100 15/15 [==============================] - 2s 134ms/step - loss: 1.4346 - accuracy: 0.5413 - val_loss: 3.9738 - val_accuracy: 0.0853 Epoch 16/100 15/15 [==============================] - 2s 135ms/step - loss: 1.5008 - accuracy: 0.5127 - val_loss: 3.5753 - val_accuracy: 0.1029 Epoch 17/100 15/15 [==============================] - 2s 136ms/step - loss: 1.3924 - accuracy: 0.5563 - val_loss: 3.4387 - val_accuracy: 0.1588 Epoch 18/100 15/15 [==============================] - 2s 137ms/step - loss: 1.2455 - accuracy: 0.5805 - val_loss: 3.3594 - val_accuracy: 0.2059 Epoch 19/100 15/15 [==============================] - 2s 139ms/step - loss: 1.2719 - accuracy: 0.5582 - val_loss: 3.5453 - val_accuracy: 0.1500 Epoch 20/100 15/15 [==============================] - 2s 137ms/step - loss: 1.3197 - accuracy: 0.5634 - val_loss: 3.1898 - val_accuracy: 0.1971 Epoch 21/100 15/15 [==============================] - 2s 137ms/step - loss: 1.2586 - accuracy: 0.5803 - val_loss: 2.8587 - val_accuracy: 0.2324 Epoch 22/100 15/15 [==============================] - 2s 139ms/step - loss: 1.2152 - accuracy: 0.5943 - val_loss: 2.9865 - val_accuracy: 0.2265 Epoch 23/100 15/15 [==============================] - 2s 138ms/step - loss: 1.2592 - accuracy: 0.5697 - val_loss: 2.5172 - val_accuracy: 0.3059 Epoch 24/100 15/15 [==============================] - 2s 135ms/step - loss: 1.1461 - accuracy: 0.6181 - val_loss: 2.4704 - val_accuracy: 0.2941 Epoch 25/100 15/15 [==============================] - 2s 135ms/step - loss: 1.1085 - accuracy: 0.6289 - val_loss: 2.4467 - val_accuracy: 0.3029 Epoch 26/100 15/15 [==============================] - 2s 135ms/step - loss: 1.1656 - accuracy: 0.6024 - val_loss: 2.4243 - val_accuracy: 0.2941 Epoch 27/100 15/15 [==============================] - 2s 135ms/step - loss: 1.0927 - accuracy: 0.6307 - val_loss: 2.0611 - val_accuracy: 0.3500 Epoch 28/100 15/15 [==============================] - 2s 135ms/step - loss: 1.1269 - accuracy: 0.6057 - val_loss: 2.0753 - val_accuracy: 0.3588 Epoch 29/100 15/15 [==============================] - 2s 135ms/step - loss: 1.0104 - accuracy: 0.6635 - val_loss: 2.1092 - val_accuracy: 0.3529 Epoch 30/100 15/15 [==============================] - 2s 135ms/step - loss: 1.1742 - accuracy: 0.6131 - val_loss: 1.8203 - val_accuracy: 0.4088 Epoch 31/100 15/15 [==============================] - 2s 135ms/step - loss: 1.0229 - accuracy: 0.6390 - val_loss: 1.6365 - val_accuracy: 0.4471 Epoch 32/100 15/15 [==============================] - 2s 135ms/step - loss: 1.0550 - accuracy: 0.6501 - val_loss: 1.7137 - val_accuracy: 0.4265 Epoch 33/100 15/15 [==============================] - 2s 134ms/step - loss: 1.0414 - accuracy: 0.6519 - val_loss: 1.6315 - val_accuracy: 0.4647 Epoch 34/100 15/15 [==============================] - 2s 137ms/step - loss: 1.0395 - accuracy: 0.6668 - val_loss: 1.4633 - val_accuracy: 0.5088 Epoch 35/100 15/15 [==============================] - 2s 141ms/step - loss: 1.0582 - accuracy: 0.6355 - val_loss: 1.6423 - val_accuracy: 0.4441 Epoch 36/100 15/15 [==============================] - 2s 137ms/step - loss: 0.9503 - accuracy: 0.6792 - val_loss: 1.6750 - val_accuracy: 0.4500 Epoch 37/100 15/15 [==============================] - 2s 136ms/step - loss: 0.9566 - accuracy: 0.6633 - val_loss: 1.4615 - val_accuracy: 0.5176 Epoch 38/100 15/15 [==============================] - 2s 137ms/step - loss: 0.9583 - accuracy: 0.6969 - val_loss: 1.5130 - val_accuracy: 0.5118 Epoch 39/100 15/15 [==============================] - 2s 138ms/step - loss: 0.9441 - accuracy: 0.6733 - val_loss: 1.4120 - val_accuracy: 0.5471 Epoch 40/100 15/15 [==============================] - 2s 135ms/step - loss: 0.9422 - accuracy: 0.6853 - val_loss: 1.4001 - val_accuracy: 0.5676 Epoch 41/100 15/15 [==============================] - 2s 134ms/step - loss: 1.0586 - accuracy: 0.6496 - val_loss: 1.2586 - val_accuracy: 0.6176 Epoch 42/100 15/15 [==============================] - 2s 134ms/step - loss: 0.9744 - accuracy: 0.6507 - val_loss: 1.2267 - val_accuracy: 0.6412 Epoch 43/100 15/15 [==============================] - 2s 134ms/step - loss: 0.9584 - accuracy: 0.6580 - val_loss: 1.3615 - val_accuracy: 0.6118 Epoch 44/100 15/15 [==============================] - 2s 134ms/step - loss: 0.8866 - accuracy: 0.7053 - val_loss: 1.3605 - val_accuracy: 0.5882 Epoch 45/100 15/15 [==============================] - 2s 134ms/step - loss: 0.9902 - accuracy: 0.6587 - val_loss: 1.3908 - val_accuracy: 0.5971 Epoch 46/100 15/15 [==============================] - 2s 135ms/step - loss: 0.9166 - accuracy: 0.6929 - val_loss: 1.2962 - val_accuracy: 0.6412 Epoch 47/100 15/15 [==============================] - 2s 135ms/step - loss: 0.9265 - accuracy: 0.6790 - val_loss: 1.3954 - val_accuracy: 0.6294 Epoch 48/100 15/15 [==============================] - 2s 134ms/step - loss: 0.8822 - accuracy: 0.6935 - val_loss: 1.3713 - val_accuracy: 0.6353 Epoch 49/100 15/15 [==============================] - 2s 136ms/step - loss: 0.9113 - accuracy: 0.6933 - val_loss: 1.5242 - val_accuracy: 0.6088 Epoch 50/100 15/15 [==============================] - 2s 138ms/step - loss: 0.9711 - accuracy: 0.6546 - val_loss: 1.2551 - val_accuracy: 0.6735 Epoch 51/100 15/15 [==============================] - 2s 135ms/step - loss: 0.9279 - accuracy: 0.6743 - val_loss: 1.5381 - val_accuracy: 0.6000 Epoch 52/100 15/15 [==============================] - 2s 138ms/step - loss: 0.9004 - accuracy: 0.6965 - val_loss: 1.3691 - val_accuracy: 0.6441 Epoch 53/100 15/15 [==============================] - 2s 134ms/step - loss: 0.9037 - accuracy: 0.7001 - val_loss: 1.3164 - val_accuracy: 0.6559 Epoch 54/100 15/15 [==============================] - 2s 135ms/step - loss: 0.8371 - accuracy: 0.7199 - val_loss: 1.3969 - val_accuracy: 0.6265 Epoch 55/100 15/15 [==============================] - 2s 135ms/step - loss: 0.8926 - accuracy: 0.6811 - val_loss: 1.4008 - val_accuracy: 0.6471 Epoch 56/100 15/15 [==============================] - 2s 135ms/step - loss: 0.8771 - accuracy: 0.7195 - val_loss: 1.3869 - val_accuracy: 0.6559 Epoch 57/100 15/15 [==============================] - 2s 135ms/step - loss: 0.8108 - accuracy: 0.7168 - val_loss: 1.2751 - val_accuracy: 0.6676 Epoch 58/100 15/15 [==============================] - 2s 135ms/step - loss: 0.7589 - accuracy: 0.7346 - val_loss: 1.3941 - val_accuracy: 0.6382 Epoch 59/100 15/15 [==============================] - 2s 134ms/step - loss: 0.7820 - accuracy: 0.7353 - val_loss: 1.6379 - val_accuracy: 0.5794 Epoch 60/100 15/15 [==============================] - 2s 134ms/step - loss: 0.8502 - accuracy: 0.6937 - val_loss: 1.2935 - val_accuracy: 0.6559 Epoch 61/100 15/15 [==============================] - 2s 134ms/step - loss: 0.7830 - accuracy: 0.7412 - val_loss: 1.3811 - val_accuracy: 0.6441 Epoch 62/100 15/15 [==============================] - 2s 134ms/step - loss: 0.8258 - accuracy: 0.6980 - val_loss: 1.5782 - val_accuracy: 0.6000 Epoch 63/100 15/15 [==============================] - 2s 141ms/step - loss: 0.7687 - accuracy: 0.7314 - val_loss: 1.4918 - val_accuracy: 0.6324 Epoch 64/100 15/15 [==============================] - 2s 150ms/step - loss: 0.8482 - accuracy: 0.7134 - val_loss: 1.5208 - val_accuracy: 0.6559 Epoch 65/100 15/15 [==============================] - 2s 136ms/step - loss: 0.7751 - accuracy: 0.7427 - val_loss: 1.4957 - val_accuracy: 0.6441 Epoch 66/100 15/15 [==============================] - 2s 140ms/step - loss: 0.7245 - accuracy: 0.7595 - val_loss: 1.3448 - val_accuracy: 0.6618 Epoch 67/100 15/15 [==============================] - 2s 136ms/step - loss: 0.7404 - accuracy: 0.7433 - val_loss: 1.4025 - val_accuracy: 0.6500 Epoch 68/100 15/15 [==============================] - 2s 138ms/step - loss: 0.7703 - accuracy: 0.7345 - val_loss: 1.2842 - val_accuracy: 0.6941 Epoch 69/100 15/15 [==============================] - 2s 148ms/step - loss: 0.7883 - accuracy: 0.7083 - val_loss: 1.4227 - val_accuracy: 0.6353 Epoch 70/100 15/15 [==============================] - 2s 154ms/step - loss: 0.7302 - accuracy: 0.7383 - val_loss: 1.3693 - val_accuracy: 0.6735 Epoch 71/100 15/15 [==============================] - 2s 147ms/step - loss: 0.7753 - accuracy: 0.7351 - val_loss: 1.4273 - val_accuracy: 0.6382 Epoch 72/100 15/15 [==============================] - 2s 140ms/step - loss: 0.7301 - accuracy: 0.7618 - val_loss: 1.3834 - val_accuracy: 0.6618 Epoch 73/100 15/15 [==============================] - 2s 142ms/step - loss: 0.7797 - accuracy: 0.7380 - val_loss: 1.4184 - val_accuracy: 0.6706 Epoch 74/100 15/15 [==============================] - 2s 141ms/step - loss: 0.7000 - accuracy: 0.7733 - val_loss: 1.4157 - val_accuracy: 0.6618 Epoch 75/100 15/15 [==============================] - 2s 135ms/step - loss: 0.7822 - accuracy: 0.7364 - val_loss: 1.5064 - val_accuracy: 0.6324 Epoch 76/100 15/15 [==============================] - 2s 141ms/step - loss: 0.6704 - accuracy: 0.7659 - val_loss: 1.4767 - val_accuracy: 0.6588 Epoch 77/100 15/15 [==============================] - 2s 136ms/step - loss: 0.7181 - accuracy: 0.7581 - val_loss: 1.6558 - val_accuracy: 0.6206 Epoch 78/100 15/15 [==============================] - 2s 139ms/step - loss: 0.7885 - accuracy: 0.7130 - val_loss: 1.3720 - val_accuracy: 0.6706 Epoch 79/100 15/15 [==============================] - 2s 135ms/step - loss: 0.7456 - accuracy: 0.7480 - val_loss: 1.3569 - val_accuracy: 0.6676 Epoch 80/100 15/15 [==============================] - 2s 135ms/step - loss: 0.6948 - accuracy: 0.7651 - val_loss: 1.3312 - val_accuracy: 0.6794 Epoch 81/100 15/15 [==============================] - 2s 138ms/step - loss: 0.6994 - accuracy: 0.7673 - val_loss: 1.5909 - val_accuracy: 0.6412 Epoch 82/100 15/15 [==============================] - 2s 136ms/step - loss: 0.7452 - accuracy: 0.7415 - val_loss: 1.3472 - val_accuracy: 0.6706 Epoch 83/100 15/15 [==============================] - 2s 138ms/step - loss: 0.6759 - accuracy: 0.7584 - val_loss: 1.6702 - val_accuracy: 0.6118 Epoch 84/100 15/15 [==============================] - 2s 138ms/step - loss: 0.6271 - accuracy: 0.7601 - val_loss: 1.3936 - val_accuracy: 0.6529 Epoch 85/100 15/15 [==============================] - 2s 134ms/step - loss: 0.6318 - accuracy: 0.7851 - val_loss: 1.3989 - val_accuracy: 0.6706 Epoch 86/100 15/15 [==============================] - 2s 135ms/step - loss: 0.6957 - accuracy: 0.7633 - val_loss: 1.2167 - val_accuracy: 0.7029 Epoch 87/100 15/15 [==============================] - 2s 142ms/step - loss: 0.6994 - accuracy: 0.7720 - val_loss: 1.3380 - val_accuracy: 0.6706 Epoch 88/100 15/15 [==============================] - 2s 136ms/step - loss: 0.6666 - accuracy: 0.7662 - val_loss: 1.3656 - val_accuracy: 0.6559 Epoch 89/100 15/15 [==============================] - 2s 135ms/step - loss: 0.7065 - accuracy: 0.7554 - val_loss: 1.3882 - val_accuracy: 0.6794 Epoch 90/100 15/15 [==============================] - 2s 140ms/step - loss: 0.6378 - accuracy: 0.7741 - val_loss: 1.4967 - val_accuracy: 0.6235 Epoch 91/100 15/15 [==============================] - 2s 141ms/step - loss: 0.6500 - accuracy: 0.7674 - val_loss: 1.3248 - val_accuracy: 0.6853 Epoch 92/100 15/15 [==============================] - 2s 144ms/step - loss: 0.6552 - accuracy: 0.7628 - val_loss: 1.3370 - val_accuracy: 0.6765 Epoch 93/100 15/15 [==============================] - 2s 136ms/step - loss: 0.5937 - accuracy: 0.7848 - val_loss: 1.3386 - val_accuracy: 0.6676 Epoch 94/100 15/15 [==============================] - 2s 137ms/step - loss: 0.6792 - accuracy: 0.7791 - val_loss: 1.9461 - val_accuracy: 0.5824 Epoch 95/100 15/15 [==============================] - 2s 148ms/step - loss: 0.6887 - accuracy: 0.7543 - val_loss: 1.5584 - val_accuracy: 0.6382 Epoch 96/100 15/15 [==============================] - 2s 136ms/step - loss: 0.5968 - accuracy: 0.7966 - val_loss: 1.8105 - val_accuracy: 0.6176 Epoch 97/100 15/15 [==============================] - 2s 135ms/step - loss: 0.5653 - accuracy: 0.7941 - val_loss: 1.3839 - val_accuracy: 0.6676 Epoch 98/100 15/15 [==============================] - 2s 135ms/step - loss: 0.5649 - accuracy: 0.7997 - val_loss: 1.3729 - val_accuracy: 0.6882 Epoch 99/100 15/15 [==============================] - 2s 135ms/step - loss: 0.5878 - accuracy: 0.7927 - val_loss: 1.4815 - val_accuracy: 0.6618 Epoch 100/100 15/15 [==============================] - 2s 135ms/step - loss: 0.6179 - accuracy: 0.7935 - val_loss: 1.2644 - val_accuracy: 0.7059
predictions=net.predict(test_x, batch_size=config.BATCH_SIZE)
print(classification_report(test_y.argmax(axis=1),predictions.argmax(axis=1), target_names=classes))
precision recall f1-score support
0 0.32 0.47 0.38 15
1 0.54 0.95 0.69 20
10 0.84 0.78 0.81 27
11 0.75 0.60 0.67 20
12 0.56 0.88 0.68 16
13 0.80 0.44 0.57 18
14 0.89 0.70 0.78 23
15 0.87 0.65 0.74 20
16 1.00 0.53 0.69 19
2 0.89 0.44 0.59 18
3 1.00 0.77 0.87 22
4 0.51 0.90 0.65 20
5 0.94 0.70 0.80 23
6 0.82 0.86 0.84 21
7 0.47 0.61 0.53 23
8 0.81 0.81 0.81 16
9 0.89 0.84 0.86 19
accuracy 0.71 340
macro avg 0.76 0.70 0.70 340
weighted avg 0.77 0.71 0.71 340
plot_training_metrics(Z,epochs=config.EPOCHS,inline=True)
mesh, class_name,class_index=confusion_mesh(test_y.argmax(axis=1),
predictions.argmax(axis=1),
encoded_classes=LabelEncoder().fit_transform(classes),
class_name="16",
classes=classes.tolist())
plot_confusion_mesh(mesh, class_name)
test_image, class_name=random_image(config.DATASET_PATH,class_index=config.CLASS_INDEX)
display_image(image=test_image)
class_name
'6'
preprocessed_test_image=preprocess_image(preprocessors=preprocessors,image=test_image, for_prediction=True)
test_predict=net.predict(preprocessed_test_image)
display_prediction(test_predict.argmax(axis=1),classes, image=test_image,font_scale=0.7 ,text_thickness=1)
