CNN 迁移学习 vgg16
应用场景
假如我们有一系列诉求是把图片识别成一个特定分类、比如
- 把图片分类成为猫、狗、狼等
- 把图片分类成为奔驰、宝马、奥迪
- ...
几乎很少有人从头训练网络、复用只有训练的网络参数适应新的数据集、参考transfer-learning
In practice, very few people train an entire Convolutional Network from scratch (with random initialization), because it is relatively rare to have a dataset of sufficient size. Instead, it is common to pretrain a ConvNet on a very large dataset (e.g. ImageNet, which contains 1.2 million images with 1000 categories), and then use the ConvNet either as an initialization or a fixed feature extractor for the task of interest.
prerequisite知识
. CNN卷积过程
. TensorFlow的接口
在可视化下贴上caffemodel定义可以查看网络结构、以下是vgg16前几层的参考
层数越往上激活的图片就约简单、所以更容易被共享；拿用image Net训练好1000分类的网络参数可以认为前几层几乎都是训练好的、替换最后面fc层、换成目标的分类的个数
假如我们识别的是猫狗、那么fc就两个分类、最后一层需要重新训练
代码实例
基于TensorFlow vgg16 fine tuning
卷积矩阵大小变化变化可以参考过程、
其中涉及数据预处理可以参考neural-networks-2
Mean subtraction is the most common form of preprocessing. It involves subtracting the mean across every individual feature in the data, and has the geometric interpretation of centering the cloud of data around the origin along every dimension. In numpy, this operation would be implemented as: X -= np.mean(X, axis = 0). With images specifically, for convenience it can be common to subtract a single value from all pixels (e.g. X -= np.mean(X)), or to do so separately across the three color channels.
代码如下：""" #训练好的参数http://download.tensorflow.org/models/vgg_16_2016_08_28.tar.gz 目录结构 train/ 猫/ COCO_train2014_000000005785.jpg COCO_train2014_000000015870.jpg 🐶/ val/ 猫/ 狗/ """ import argparse import os import tensorflow as tf import tensorflow.contrib.slim as slim import tensorflow.contrib.slim.nets parser = argparse.ArgumentParser() #训练数据目录 parser.add_argument('--train_dir', default='train') #测试目录 parser.add_argument('--val_dir', default='val') #初始网络参数 parser.add_argument('--model_path', default='vgg_16.ckpt', type=str) parser.add_argument('--batch_size', default=32, type=int) parser.add_argument('--num_workers', default=4, type=int) parser.add_argument('--num_epochs1', default=10, type=int) parser.add_argument('--num_epochs2', default=10, type=int) parser.add_argument('--learning_rate1', default=1e-3, type=float) parser.add_argument('--learning_rate2', default=1e-5, type=float) parser.add_argument('--dropout_keep_prob', default=0.5, type=float) parser.add_argument('--weight_decay', default=5e-4, type=float) #平化图像参数 VGG_MEAN = [123.68, 116.78, 103.94] def list_images(directory): labels = os.listdir(directory) files_and_labels = [] for label in labels: for f in os.listdir(os.path.join(directory, label)): files_and_labels.append((os.path.join(directory, label, f), label)) filenames, labels = zip(*files_and_labels) filenames = list(filenames) labels = list(labels) unique_labels = list(set(labels)) label_to_int = {} for i, label in enumerate(unique_labels): label_to_int[label] = i labels = [label_to_int[l] for l in labels] return filenames, labels def check_accuracy(sess, correct_prediction, is_training, dataset_init_op): # Initialize the correct dataset sess.run(dataset_init_op) num_correct, num_samples = 0, 0 while True: try: correct_pred = sess.run(correct_prediction, {is_training: False}) num_correct += correct_pred.sum() num_samples += correct_pred.shape[0] except tf.errors.OutOfRangeError: break acc = float(num_correct) / num_samples return acc def main(args): # 拿训练&测试文件和label train_filenames, train_labels = list_images(args.train_dir) val_filenames, val_labels = list_images(args.val_dir) num_classes = len(set(train_labels)) graph = tf.Graph() with graph.as_default(): #读图 def _parse_function(filename, label): image_string = tf.read_file(filename) image_decoded = tf.image.decode_jpeg(image_string, channels=3) image = tf.cast(image_decoded, tf.float32) smallest_side = 256.0 height, width = tf.shape(image)[0], tf.shape(image)[1] height = tf.to_float(height) width = tf.to_float(width) #缩放 scale = tf.cond(tf.greater(height, width), lambda: smallest_side / width, lambda: smallest_side / height) new_height = tf.to_int32(height * scale) new_width = tf.to_int32(width * scale) resized_image = tf.image.resize_images(image, [new_height, new_width]) # (2) return resized_image, label #均值数据处理 def training_preprocess(image, label): crop_image = tf.random_crop(image, [224, 224, 3]) # (3) flip_image = tf.image.random_flip_left_right(crop_image) # (4) means = tf.reshape(tf.constant(VGG_MEAN), [1, 1, 3]) centered_image = flip_image - means # (5) return centered_image, label # 预处理、取224*224中间区域、减平均值 def val_preprocess(image, label): crop_image = tf.image.resize_image_with_crop_or_pad(image, 224, 224) # (3) means = tf.reshape(tf.constant(VGG_MEAN), [1, 1, 3]) centered_image = crop_image - means # (4) return centered_image, label train_filenames = tf.constant(train_filenames) train_labels = tf.constant(train_labels) train_dataset = tf.contrib.data.Dataset.from_tensor_slices((train_filenames, train_labels)) train_dataset = train_dataset.map(_parse_function, num_threads=args.num_workers, output_buffer_size=args.batch_size) train_dataset = train_dataset.map(training_preprocess, num_threads=args.num_workers, output_buffer_size=args.batch_size) train_dataset = train_dataset.shuffle(buffer_size=10000) batched_train_dataset = train_dataset.batch(args.batch_size) val_filenames = tf.constant(val_filenames) val_labels = tf.constant(val_labels) val_dataset = tf.contrib.data.Dataset.from_tensor_slices((val_filenames, val_labels)) val_dataset = val_dataset.map(_parse_function, num_threads=args.num_workers, output_buffer_size=args.batch_size) val_dataset = val_dataset.map(val_preprocess, num_threads=args.num_workers, output_buffer_size=args.batch_size) batched_val_dataset = val_dataset.batch(args.batch_size) #迭代器读图&label iterator = tf.contrib.data.Iterator.from_structure(batched_train_dataset.output_types, batched_train_dataset.output_shapes) images, labels = iterator.get_next() #初始化迭代器函数 train_init_op = iterator.make_initializer(batched_train_dataset) val_init_op = iterator.make_initializer(batched_val_dataset) #传给vgg16网络、标识正向分类或者是训练网络参数 is_training = tf.placeholder(tf.bool) vgg = tf.contrib.slim.nets.vgg with slim.arg_scope(vgg.vgg_arg_scope(weight_decay=args.weight_decay)): #使用TensorFlow封装好的网络、设置输出分类个数 logits, _ = vgg.vgg_16(images, num_classes=num_classes, is_training=is_training, dropout_keep_prob=args.dropout_keep_prob) model_path = args.model_path assert(os.path.isfile(model_path)) # 加载fc8之前网络参数 variables_to_restore = tf.contrib.framework.get_variables_to_restore(exclude=['vgg_16/fc8']) init_fn = tf.contrib.framework.assign_from_checkpoint_fn(model_path, variables_to_restore) # 获取fc8初始化函数 fc8_variables = tf.contrib.framework.get_variables('vgg_16/fc8') fc8_init = tf.variables_initializer(fc8_variables) # loss叠加到tf.GraphKeys.LOSSES 结合上 tf.losses.sparse_softmax_cross_entropy(labels=labels, logits=logits) loss = tf.losses.get_total_loss() #先训练fc8这一层的参数 fc8_optimizer = tf.train.GradientDescentOptimizer(args.learning_rate1) fc8_train_op = fc8_optimizer.minimize(loss, var_list=fc8_variables) # 然后再去整体训练 full_optimizer = tf.train.GradientDescentOptimizer(args.learning_rate2) full_train_op = full_optimizer.minimize(loss) # 评估模型 prediction = tf.to_int32(tf.argmax(logits, 1)) correct_prediction = tf.equal(prediction, labels) accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32)) tf.get_default_graph().finalize() with tf.Session(graph=graph) as sess: #加载conv1-fc7的参数 init_fn(sess) #初始化fc的参数 sess.run(fc8_init) #迭代 for epoch in range(args.num_epochs1): sess.run(train_init_op) while True: try: #文件和label已经在迭代器中 _ = sess.run(fc8_train_op, {is_training: True}) except tf.errors.OutOfRangeError: break train_acc = check_accuracy(sess, correct_prediction, is_training, train_init_op) val_acc = check_accuracy(sess, correct_prediction, is_training, val_init_op) # 整体训练 for epoch in range(args.num_epochs2): print('Starting epoch %d / %d' % (epoch + 1, args.num_epochs2)) sess.run(train_init_op) while True: try: _ = sess.run(full_train_op, {is_training: True}) except tf.errors.OutOfRangeError: break train_acc = check_accuracy(sess, correct_prediction, is_training, train_init_op) val_acc = check_accuracy(sess, correct_prediction, is_training, val_init_op) print('Train accuracy: %f' % train_acc) print('Val accuracy: %f\n' % val_acc) if __name__ == '__main__': args = parser.parse_args() main(args)
这个例子不需要GPU的支持、在osx就可以跑
def vgg_16(inputs, num_classes=1000, is_training=True, dropout_keep_prob=0.5, spatial_squeeze=True, scope='vgg_16'): """Oxford Net VGG 16-Layers version D Example. Note: All the fully_connected layers have been transformed to conv2d layers. To use in classification mode, resize input to 224x224. Args: inputs: a tensor of size [batch_size, height, width, channels]. num_classes: number of predicted classes. is_training: whether or not the model is being trained. dropout_keep_prob: the probability that activations are kept in the dropout layers during training. spatial_squeeze: whether or not should squeeze the spatial dimensions of the outputs. Useful to remove unnecessary dimensions for classification. scope: Optional scope for the variables. Returns: the last op containing the log predictions and end_points dict.
结论
通常工程同学不会设计新的网络结构、甚至很少大改一个网络机构、但是理解网络结构、loss渐进方式有利于迁移学习、用到特定的场景
