由於語法渲染問題而影響閱讀體驗， 請移步博客閱讀～
本文GitPage地址

Transfer_Learning

  1. ##!/usr/local/bin/python3.6
  3. from tensorflow.python.keras.applications import ResNet50
  4. from tensorflow.python.keras.models import Sequential
  5. from tensorflow.python.keras.layers import Dense, Flatten, GlobalAveragePooling2D
  7. num_classes = 2
  8. resnet_weights_path = 'resnet50_weights_tf_dim_ordering_tf_kernels_notop.h5'
  11. my_new_model = Sequential()
  12. my_new_model.add(ResNet50(include_top=False, pooling='avg', weights=resnet_weights_path))
  13. my_new_model.add(Dense(num_classes, activation='softmax'))
  16. my_new_model.layers[0].trainable = False
  18. my_new_model.compile(optimizer='sgd', loss='categorical_crossentropy', metrics=['accuracy'])
  20. ######## Fit Model ###
  21. from tensorflow.python.keras.applications.resnet50 import preprocess_input
  22. from tensorflow.python.keras.preprocessing.image import ImageDataGenerator
  24. image_size = 224
  25. data_generator = ImageDataGenerator(preprocessing_function=preprocess_input)
  27. train_generator = data_generator.flow_from_directory(
  28.         './train',
  29.         target_size=(image_size, image_size),
  30.         batch_size=24,
  31.         class_mode='categorical')
  33. validation_generator = data_generator.flow_from_directory(
  34.         './val',
  35.         target_size=(image_size, image_size),
  36.         class_mode='categorical')
  38. my_new_model.fit_generator(
  39.         train_generator,
  40.         steps_per_epoch=3,
  41.         validation_data=validation_generator,
  42.         validation_steps=1)

Enjoy~

本文由Python腳本GitHub/語雀自動更新

由於語法渲染問題而影響閱讀體驗， 請移步博客閱讀～
本文GitPage地址

GitHub: Karobben
Blog:Karobben
BiliBili:史上最不正經的生物狗