初窥Tensorflow Object Detection API 源码之（1.2）FeatureExtractor.Config

type

batch_norm_trainable

first_stage_features_stride

message FasterRcnnFeatureExtractor {
// Type of Faster R-CNN model (e.g., 'faster_rcnn_resnet101';
// See builders/model_builder.py for expected types).
optional string type = 1;

// Output stride of extracted RPN feature map.
optional int32 first_stage_features_stride = 2 [default=16];

// Whether to update batch norm parameters during training or not.
// When training with a relative large batch size (e.g. 8), it could be
// desirable to enable batch norm update.
optional bool batch_norm_trainable = 3 [default=false];
}

type

string类型

值：faster_rcnn_resnet101 等

batch_norm_trainable

默认false

first_stage_features_stride

FasterRCNNResnetV1FeatureExtractor的

__init__

获得了该配置，然后传给了父类FasterRCNNFeatureExtractor的

__init__

if first_stage_features_stride != 8 and first_stage_features_stride != 16:
raise ValueError('`first_stage_features_stride` must be 8 or 16.')

注意：该配置的值只能是8或16，否则报错

在FasterRCNNFeatureExtractor的

__init__

中，又观察到

self._first_stage_features_stride = first_stage_features_stride

该变量被赋值给了私有变量，然后我在该类和它的派生类中搜索

_first_stage_features_stride

，发现了在

_extract_proposal_features

这样一段

with tf.variable_scope(
self._architecture, reuse=self._reuse_weights) as var_scope:
_, activations = self._resnet_model(
preprocessed_inputs,
num_classes=None,
is_training=self._train_batch_norm,
global_pool=False,
output_stride=self._first_stage_features_stride,
spatial_squeeze=False,
scope=var_scope)

值又被传给了

_resnet_model

的

output_stride

，好复杂有木有？面向对象就要这么绕，习惯就好。这里的

_resnet_model

，猜也能猜到，是resnet的函数

resnet_v1.py

中找到了

output_stride

的解释

output_stride: If None, then the output will be computed at the nominal

network stride. If output_stride is not None, it specifies the requested

ratio of input to output spatial resolution.

意思是定义了从输入到输出的分辨率比

但此时一想，不对啊，featureExtractor只使用了resnet的前3个block的输出，然后替换上了自己的block4,为什么输出分辨率比没有变化

这里有两个注意点：

1. resnet50,101,152,200等网络的block4的stride都为1，且FeatureExtractor的block4的stride亦为1，所以替换不影响stride；

2. resnet默认对输入阶段就有conv2d、max_pool2d，使得output_stride/=4

if include_root_block:
if output_stride is not None:
if output_stride % 4 != 0:
raise ValueError('The output_stride needs to be a multiple of 4.')
output_stride /= 4
net = resnet_utils.conv2d_same(net, 64, 7, stride=2, scope='conv1')
net = slim.max_pool2d(net, [3, 3], stride=2, scope='pool1')

include_root_block默认为true