DeepLearning.AI TensorFlow Developer Course 1-3

Posted on | In

[Coursera] DeepLearning.AI TensorFlow Developer ( Tensorflow In Practice ) 강의 정리
Course 1 : Introduction to TensorFlow for Artificial Intelligence, Machine Learning, and Deep Learning

  • Week 1 - A New Programming Paradigm
  • Week 2 - Introduction to Computer Vision
  • Week 3 - Enhancing Vision with Convolutional Neural Networks
  • Week 4 - Using Real-world Images

Course 2 : Convolutional Neural Networks in TensorFlow
Course 3 : Natural Language Processing in TensorFlow
Course 4 : Sequences, Time Series and Prediction

Week 3 - Enhancing Vision with Convolutional Neural Networks

fashion mnist 데이터로 이어서 강의 진행
앞서 사용한 기존 모델 코드

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
import numpy as np
import tensorflow as tf

mnist = tf.keras.datasets.fashion_mnist
(x_train, y_train), (x_test, y_test) = mnist.load_data()

x_train, x_test = x_train / 255.0, x_test / 255.0

model = tf.keras.models.Sequential([
tf.keras.layers.Flatten(input_shape=(28, 28)),
tf.keras.layers.Dense(512, activation=tf.nn.relu),
tf.keras.layers.Dense(10, activation=tf.nn.softmax)
])
model.compile(optimizer=tf.optimizers.Adam(),
loss='sparse_categorical_crossentropy',
metrics=['accuracy'])
model.fit(x_train, y_train, epochs=10)

Implementing convolutional layers

기존 모델

1
2
3
4
5
model = tf.keras.models.Sequential([
tf.keras.layers.Flatten(),
tf.keras.layers.Dense(128, activation=tf.nn.relu),
tf.keras.layers.Dense(10, activation=tf.nn.softmax)
])

convolutional layer 사용하는 모델
마지막 세 줄은 기존 모델과 동일
tf.keras.layers.Flatten()
tf.keras.layers.Dense(128, activation='relu')
tf.keras.layers.Dense(10, activation='softmax')

1
2
3
4
5
6
7
8
9
model = tf.keras.models.Sequential([
tf.keras.layers.Conv2D(64, (3, 3), activation='relu', input_shape=(28,28,1)),
tf.keras.layers.MaxPooling2D(2, 2),
tf.keras.layers.Conv2D(64, (3, 3), activation='relu'),
tf.keras.layers.MaxPooling2D(2, 2),
tf.keras.layers.Flatten(),
tf.keras.layers.Dense(128, activation='relu'),
tf.keras.layers.Dense(10, activation='softmax')
])

tf.keras.layers.Conv2D(64, (3, 3), activation='relu', input_shape=(28,28,1))
첫 번째 convolution 지정
karas에 64개 필터 생성을 요청 - 필터는 3 x 3이고, actiovation은 relu, 입력형태는 28 x 28이다.
추가한 1은 색 깊이를 의미한다.

Implementing pooling layers

1
2
3
4
5
6
7
8
9
10
11
model = tf.keras.models.Sequential([
tf.keras.layers.Conv2D(64, (3, 3), activation='relu', input_shape=(28,28,1)),
tf.keras.layers.MaxPooling2D(2, 2),
tf.keras.layers.Conv2D(64, (3, 3), activation='relu'),
tf.keras.layers.MaxPooling2D(2, 2),
tf.keras.layers.Flatten(),
tf.keras.layers.Dense(128, activation='relu'),
tf.keras.layers.Dense(10, activation='softmax')
])

model.summary()

tf.keras.layers.MaxPooling2D(2, 2) 2 x 2 maxpooling layer
4 픽셀마다 가장 큰 수를 사용

model.summary() : 모델의 레이어 검사하고 합성곱을 통해 이동 경로 확인
Output :

Model: "sequential_3"
_________________________________________________________________ 
Layer (type)                    Output Shape          Param # 
================================================================= 
conv2d_6 (Conv2D)               (None, 26, 26, 64)    640 
_________________________________________________________________ 
max_pooling2d_4 (MaxPooling2 	(None, 13, 13, 64)    0 
_________________________________________________________________ 
conv2d_7 (Conv2D)               (None, 11, 11, 64)    36928 
_________________________________________________________________ 
max_pooling2d_5 (MaxPooling2 	(None, 5, 5, 64)      0 
_________________________________________________________________ 
flatten_3 (Flatten)             (None, 1600)          0 
_________________________________________________________________ 
dense_6 (Dense)                 (None, 128)           204928 
_________________________________________________________________ 
dense_7 (Dense)                 (None, 10)            1290 
================================================================= 
Total params: 243,786 
Trainable params: 243,786 
Non-trainable params: 0 
_________________________________________________________________

처음 26, 26은 필터가 3 x 3으로 모서리 부분은 지날 수 없으므로, x와 y의 픽셀을 하나씩 제거
maxpooling에 의해 이미지가 축소
다음 convolutional layer도 x와 y 한 픽셀씩 제거
마지막 maxpooling을 거치면, 5x5의 64개 이미지 형성

Improving the Fashion classifier with convolutions

filter - 이미지 전달하여 정보의 양을 줄이고, neural network가 한 종류의 이미지를 다른 종류와 구별할 수 있는 특징 추출
pooling - 정보를 더 쉽게 관리하기 위해 압축하는 방법


https://www.coursera.org/professional-certificates/tensorflow-in-practice

0%