这里后期会补上中间的VGGNet和Inception，这里先附上ResNet代码，后期进行补充噢！
# coding=utf-8import tensorflow as tfimport collectionsimport timefrom datetime import datetimeimport math# contrib.slim中的一些功能和组件可以大大减少设计Inception Net的代码量slim = tf.contrib.slim# 使用collections.namedtuple设计Reset基本Block模块组的named tupe，并创建Block的类，但只包含数据结构，不包含具体方法。class Block(collections.namedtuple('Block', ['scope', 'unit_fn', 'args'])): 'A named tuple describiing a ResNet block.'"""定义一个降采样subsample的方法，参数包括input(输入)，factor（采样因子）和scope。这个函数也非常简单，如果factor为1,则不做修改直接返回input;如果不为1,则使用slim.max_pool2d最大池化来实现，通过1×1的池化尺寸，stride作为步长，即可实现降采样。"""def subsample(inputs, factor, scope=None): if factor == 1: return inputs else: return slim.max_pool2d(inputs, [1, 1], stride=factor, scope=scope)# 创建一个卷积层def conv2d_same(inputs, num_outputs, kernel_size, stride, scope=None): if stride == 1: return slim.conv2d(inputs, num_outputs, kernel_size, stride=1, padding='SAME', scope=scope) else: pad_total = kernel_size - 1 pad_beg = pad_total // 2 pad_end = pad_total - pad_beg inputs = tf.pad(inputs, [[0, 0], [pad_beg, pad_end], [pad_beg, pad_end], [0, 0]]) return slim.conv2d(inputs, num_outputs, kernel_size, stride=stride, padding='VALID', scope=scope)# 定义堆叠Block的函数，参数中的net即为输入，blocks是之前定义的Block的class的列表，# 而out_collections则是用来收集各个end_points的collections。@slim.add_arg_scopedef stack_blocks_dense(net, blocks, outputs_collections): for block in blocks: with tf.variable_scope(block.scope, 'block', [net]) as sc: for i, unit in enumerate(block.args): with tf.variable_scope('unit_%d' % (i+1), values=[net]): unit_depth, unit.depth_bottleneck, unit_stride = unit net = block.unit_fn(net, depth=unit_depth, depth_bottleneck=unit_depth_bottleneck, stride=unit_stride) net = slim.utils.collect_named_outputs(outputs_collections, sc.name, net) return net# 创建ReNet通用的arg_scope，关于arg_scope，用来定义某些函数的参数def resnet_arg_scope(is_training=True, weight_decay=0.0001, batch_norm_decay=0.997, batch_norm_epsilon=1e-5, batch_norm_scale=True): batch_norm_params = { 'is_training': is_training, 'decay': batch_norm_decay, 'epsilon': batch_norm_epsilon, 'scale': batch_norm_scale 'updates_collections': tf.GraphKeys.UPDATE_OPS, } with slim.arg_scope([slim.conv2d], weights_regularizer=slim.l2_regularizer(weight_decay), weights_initializer=slim.variance_scaling_initializer(), activation_fn=tf.nn.relu, normalizer_fn=slim.batch_norm_parames): with slim.arg_scope([slim.batch_norm], **batch_norm_params): with slim.arg_scope([slim.max_pool2d], padding='SAME') as arg_sc: return arg_sc# 定义核心的bottleneck残差学习单元，它是ResNet V2的沦文中提到的Full PReactivation Residual Unit的一o个变种# 它和Resnet V1中的残差学习单元的主要区别有两点：# 一是：在每一层前都用了Batch Normalization，# 二是：对输入进行preactivation，而不是在卷积进行激活函数处理。@slim.add_arg_scopedef bottleneck(inputs, depth, depth_bottleneck, stride, outputs_collections=None, scope=None): with tf.variable_scope(scope, 'bottleneck_v2', [inputs]) as sc: depth_in = slim.utils.last_dimension(inputs.get_shape(), min_rank=4) preact = slim.batch_norm(inputs, activation_fn=tf.nn.relu, scope='preact') if depth == depth_in: shortcut = subsample(inputs, stride, 'shortcut') else: shortcut = slim.conv2d(preact, depth, [1, 1], stride=stride, normalizer_fn=None, activation_fn=None, scope='shortcut') residual = slim.conv2d(preact, depth_bottleneck, [1, 1], stride=1, scope='conv1') residual = conv2d_same(residual, depth_bottleneck, 3, stride=1, normalize_fn=None, activation_fn=None, scope='conv2') residual = slim.conv2d(residual, depth, [1, 1], stride=1, normalizer_fn=None, activation_fn=None, scope='conv3') output = shortcut +residual return slim.utils.collect_named_outputs(outputs_collections, sc.name, output)# 定义生成ResetNet V2的主函数，只要预先定义好网络的残差学习模块组blocks，它就可以生成对应的完整的ResNetdef resnet_v2(inputs, blocks, num_classes=None, global_pool=True, include_root_block=True, reuse=None, scope=None): with tf.variable_scope(scope, 'resnet_v2', [inputs], reuse=reuse) as sc: end_point_collection = sc.original_name_scope + '_end_points' with slim.arg_scope([slim.conv2d, bottleneck, stack_blocks_dense], outputs_collections=end_point_collection): net = inputs if include_root_block: with slim.arg_scope([slim.conv2d], activation_fn=None, normalizer_fn=None): net = conv2d_same(net, 64, 7, stride=2, scope='conv1') net = slim.max_pool2d(net, [3, 3], stride=2, scope='pool1') net = stack_blocks_dense(net, blocks) net = slim.batch_norm(net, activation_fn=tf.nn.relu, scope='postnorm') if global_pool: net = tf.reduce_mean(net, [1, 2], name='pool5', keep_dims=True) if num_classes is not None: net = slim.conv2d(net, num_classes, [1, 1], activation_fn=None, normalizer_fn=None, scope='logits') end_points = slim.utils.convert_collection_to_dict(end_point_collection) if num_classes is not None: end_points['predictions'] = slim.softmax(net, scope='prediction') return net, end_points# 推荐几个不同深度的ResNet网络配置，来设计层数分布为50, 101, 152和200的ResNetdef resnet_v2_50(inputs, num_classes=None, global_pool=True, reuse=True, scope='resnet_v2_50'): blocks = [ Block('block1', bottleneck, [(256, 64, 1)]*2 + [(256, 64, 2)], Block('block2', bottleneck, [(512, 128, 1)]*3 + [(512, 128, 2)]), Block('block3', bottleneck, [(1024, 256, 1)]*5 + [(1024, 256, 2)]), Block('block4', bottleneck, [(2048, 512, 1)]*3)] return resnet_v2(inputs, blocks, num_classes, global_pool, include_root_block=True, reuse=reuse, scope=scope)# 101层的ResNet和50层相比，主要变化就是把4个Blocks的unit的数增加到接近4倍def resnet_v2_101(inputs, num_classes=None, global_pool=True, reuse=True, scope='resnet_v2_101'): blocks = [ Block('block1', bottleneck, [(256, 64, 1)]*2 + [(256, 64, 2)]), Block('block2', bottleneck, [(512, 128, 1)]*3 + [(512, 128, 2)]), Block('block3', bottleneck, [(1024, 256, 1)]*22 + [(1024, 256, 2)]), Block('block4', bottleneck, [(2048, 512, 1)]*3)] return resnet_v2(inputs, blocks, num_classes, global_pool, include_root_block=True, reuse=reuse, scope=scope)# 然后152层的ResNet，则是将第二个Block的units数提高到8,# 将第三个Block的units数提高到36,Units数量提升的主要场所依然是第三个Block。def resnet_v2_152(inputs, num_classes=None, global_pool=True, reuse=True, scope='resnet_v2_152'): blocks = [ Block('block1', bottleneck, [(256, 64, 1)]*2 + [(256, 64, 2)]), Block('block2', bottleneck, [(512, 128, 1)]*7 + [(256, 128, 2)]), Block('block3', bottleneck, [(1024, 256, 1)]*35 + [(512, 256, 2)]), Block('block4', bottleneck, [(2048, 512, 1)]*3)] return resnet_v2(inputs, blocks, num_classes, global_pool, include_root_block=True, reuse=reuse, scop=scope)# 最后，200层的ResNet相比152层的ResNet，没有继续提升第三个Block的units数，# 而是将第二个Block的units数一下子提升到了23def resnet_v2_200(inputs, num_classes=None, global_pool=True, reuse=True, scope='resnet_v2_200'): blocks = [ Block('block1', bottleneck, [(256, 64, 1)]*2 + [(256, 64, 2)]), Block('block2', bottleneck, [(512, 128, 1)]*23 + [(512, 128, 2)]), Block('block3', bottleneck, [(1024, 512, 1)]*35 + [(1024, 512, 2)]), Block('block4', bottleneck, [(2048, 512, 1)]*3)] return resnet_v2(inputs, blocks, num_classes, global_pool, include_root_block=True, reuse=reuse, scope=scope)batch_size = 32height, width = 224, 224inputs = tf.random_uniform((batch_size, height, width, 3))with slim.arg_scope(resnet_arg_scope(is_training=False)): net, end_points = resnet_v2_152(inputs, 1000)init = tf.global_variables_initializer()sess = tf.session()sess.run(init)num_batches = 100def time_tensorflow_run(session, target, info_string): num_step_burn_in = 10 total_duration = 0.0 total_duration_squared = 0.0 """ 进行num_batches + num_step_burn_in次迭代计算，使用time.time()记录时间，每次迭代通过session.run(target)执行。 在初始热身的num_step_burn_in次迭代后，每10轮迭代显示当前所需要的时间。 同时每轮total_duration和total_duration_squared累加，以便后面计算每轮耗时的均值和标准差。 """ for i in range(num_batches+num_step_burn_in): start_time = time.time() _ = session.run(target) duration = time.time() - start_time if i >= num_step_burn_in: if not i % 10: print('%s: step %d, duration = %.3f' % (datatime.now(), i-num_step_burn_in, duration)) total_duration += duration total_duration_squared += duration*duration # 在循环结束后，计算每轮迭代的平均耗时mm和标准差sd，最后将结果显示出来。 # 这样就完成来计算每轮迭代耗时的评测函数time_tensorflow_run. mn = total_duration / num_batches vr = total_duration_squared / num_batches - mn*mn sd = math.sqrt(vr) print('%s: %s across %d step, %.3f +/- %.3f sec / batch' % (datatime.now(), info_string, num_batches, mn, sd))# 主函数部分time_tensorflow_run(sess, net, "Forward")
新闻热点
疑难解答
