classification

import tensorflow as tf

data

data = tf.keras.utils.get_file('flower_photos',
                             'https://storage.googleapis.com/download.tensorflow.org/example_images/flower_photos.tgz',
                              untar=True
                            )

data

'C:\\Users\\drogpard\\.keras\\datasets\\flower_photos'

path

import os

os.sep

'\\'

os.listdir(data)

['daisy', 'dandelion', 'LICENSE.txt', 'roses', 'sunflowers', 'tulips']

import pathlib

import glob

p = pathlib.Path(data)

p

WindowsPath('C:/Users/drogpard/.keras/datasets/flower_photos')

img = list(p.glob('*/*.jpg'))

# img # classification 용으로는 direcotry별로 이미지 저장 - 관례상 (편의 기능 사용 가능)

pipeline (holdout)

# 1
training = tf.keras.preprocessing.image_dataset_from_directory(p, 
                                                               subset='training', 
                                                               batch_size=64,
                                                               validation_split=0.3, 
                                                               seed=41) # tf.data.Dataset
validation = tf.keras.preprocessing.image_dataset_from_directory(p, 
                                                                 subset='validation', 
                                                                 validation_split=0.3,
                                                                 seed=41) # tf.data.Dataset

Found 3670 files belonging to 5 classes.
Using 2569 files for training.
Found 3670 files belonging to 5 classes.
Using 1101 files for validation.

batch

training

<BatchDataset element_spec=(TensorSpec(shape=(None, 256, 256, 3), dtype=tf.float32, name=None), TensorSpec(shape=(None,), dtype=tf.int32, name=None))>

preprocessing

training_ = training.map(lambda x,y : (x/255,y)).cache().prefetch(buffer_size=tf.data.AUTOTUNE)
validation_ = validation.map(lambda x,y : (x/255,y)).cache().prefetch(buffer_size=tf.data.AUTOTUNE)

model ( > Transfer learning)

input_ = tf.keras.Input((256,256,3))

x = tf.keras.layers.Conv2D(64, 3)(input_)
x = tf.keras.layers.BatchNormalization()(input_)
x = tf.keras.layers.ReLU()(x)
x = tf.keras.layers.MaxPool2D(3,2)(x)
x = tf.keras.layers.Conv2D(128,3)(x)
x = tf.keras.layers.BatchNormalization()(x)
x = tf.keras.layers.ReLU()(x)
x = tf.keras.layers.Conv2D(256,3)(x)
x = tf.keras.layers.BatchNormalization()(x)
x = tf.keras.layers.ReLU()(x)
x = tf.keras.layers.MaxPool2D(3,2)(x)
#x = tf.keras.layers.Flatten()(x)
x = tf.keras.layers.GlobalAvgPool2D()(x)
x = tf.keras.layers.Dense(5)(x)

model = tf.keras.Model(input_, x)

model.summary()

Model: "model"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 input_1 (InputLayer)        [(None, 256, 256, 3)]     0         
                                                                 
 batch_normalization (BatchN  (None, 256, 256, 3)      12        
 ormalization)                                                   
                                                                 
 re_lu (ReLU)                (None, 256, 256, 3)       0         
                                                                 
 max_pooling2d (MaxPooling2D  (None, 127, 127, 3)      0         
 )                                                               
                                                                 
 conv2d_1 (Conv2D)           (None, 125, 125, 128)     3584      
                                                                 
 batch_normalization_1 (Batc  (None, 125, 125, 128)    512       
 hNormalization)                                                 
                                                                 
 re_lu_1 (ReLU)              (None, 125, 125, 128)     0         
                                                                 
 conv2d_2 (Conv2D)           (None, 123, 123, 256)     295168    
                                                                 
 batch_normalization_2 (Batc  (None, 123, 123, 256)    1024      
 hNormalization)                                                 
                                                                 
 re_lu_2 (ReLU)              (None, 123, 123, 256)     0         
                                                                 
 max_pooling2d_1 (MaxPooling  (None, 61, 61, 256)      0         
 2D)                                                             
                                                                 
 global_average_pooling2d (G  (None, 256)              0         
 lobalAveragePooling2D)                                          
                                                                 
 dense (Dense)               (None, 5)                 1285      
                                                                 
=================================================================
Total params: 301,585
Trainable params: 300,811
Non-trainable params: 774
_________________________________________________________________

import tensorflow_addons as tfa

model.compile(loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
             metrics=['acc'])

train

# callbaack 

tf.keras.callbacks.EarlyStopping
tf.keras.callbacks.LearningRateScheduler
tf.keras.callbacks.ReduceLROnPlateau
tf.keras.callbacks.ModelCheckpoint

keras.callbacks.ModelCheckpoint

history = model.fit(training_, epochs=3, validation_data=validation_)

Epoch 1/3
41/41 [==============================] - 692s 17s/step - loss: 1.2940 - acc: 0.4644 - val_loss: 1.6341 - val_acc: 0.2352
Epoch 2/3
41/41 [==============================] - 720s 18s/step - loss: 1.1630 - acc: 0.5325 - val_loss: 1.7552 - val_acc: 0.2343
Epoch 3/3
41/41 [==============================] - 724s 18s/step - loss: 1.1033 - acc: 0.5613 - val_loss: 1.7961 - val_acc: 0.2343

import pandas as pd

pd.DataFrame(history.history).plot.line()

<AxesSubplot:>