个人代码阅读笔记。
# --------------------------------------------------------
# Tensorflow Faster R-CNN
# Licensed under The MIT License [see LICENSE for details]
# Written by Xinlei Chen
# --------------------------------------------------------
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import tensorflow as tf
import tensorflow.contrib.slim as slim
from tensorflow.contrib.slim import losses
from tensorflow.contrib.slim import arg_scope
import numpy as np
from layer_utils.snippets import generate_anchors_pre, generate_anchors_pre_tf
from layer_utils.proposal_layer import proposal_layer, proposal_layer_tf
from layer_utils.proposal_top_layer import proposal_top_layer, proposal_top_layer_tf
from layer_utils.anchor_target_layer import anchor_target_layer
from layer_utils.proposal_target_layer import proposal_target_layer
from utils.visualization import draw_bounding_boxes
from model.config import cfg
#主网络实例生成
class Network(object):
def __init__(self):#生成一系列实例参数，都是空的。
self._predictions = {}
self._losses = {}
self._anchor_targets = {}
self._proposal_targets = {}
self._layers = {}
self._gt_image = None
self._act_summaries = []
self._score_summaries = {}
self._train_summaries = []
self._event_summaries = {}
self._variables_to_fix = {}
def _add_gt_image(self):
# add back mean
image = self._image + cfg.PIXEL_MEANS#为什么是加上均值。
# BGR to RGB (opencv uses BGR)#通过reverse反向排序函数实现转换，用axis指定排序维度。
resized = tf.image.resize_bilinear(image, tf.to_int32(self._im_info[:2] / self._im_info[2]))
self._gt_image = tf.reverse(resized, axis=[-1])
def _add_gt_image_summary(self):#可视化groundtrue boxes
# use a customized visualization function to visualize the boxes
if self._gt_image is None:
self._add_gt_image()
image = tf.py_func(draw_bounding_boxes,
[self._gt_image, self._gt_boxes, self._im_info],
tf.float32, name="gt_boxes")
return tf.summary.image('GROUND_TRUTH', image)
def _add_act_summary(self, tensor):#直方图.很多类的方法实例生成了但是并没有使用
tf.summary.histogram('ACT/' + tensor.op.name + '/activations', tensor)
tf.summary.scalar('ACT/' + tensor.op.name + '/zero_fraction',
tf.nn.zero_fraction(tensor))
def _add_score_summary(self, key, tensor):#同上，针对得分
tf.summary.histogram('SCORE/' + tensor.op.name + '/' + key + '/scores', tensor)
def _add_train_summary(self, var):#这些可以在tensorboard里面看到。tensorboard里面的东西都是在这里定义的。可视化方法的实例。
tf.summary.histogram('TRAIN/' + var.op.name, var)
def _reshape_layer(self, bottom, num_dim, name):
input_shape = tf.shape(bottom)#tf.shape( input,name=None,out_type=tf.int32) 输出数据的维度矩阵，从外到内
with tf.variable_scope(name) as scope:#打开变量空间，获取变量
# change the channel to the caffe format
to_caffe = tf.transpose(bottom, [0, 3, 1, 2])#对输入的变量进行transpose，即维度交换。位置分别对应0 1 2 3，每个位置里面的数对应交换后的维度。
#改变维度其实数据结构变了，但实际上也就是在调用的时候，调用维度要用新的维度调用。比如a[0,1]变换维度后，还要调用该值，则调用a[1,0]
# then force it to have channel 2
reshaped = tf.reshape(to_caffe,
tf.concat(axis=0, values=[[1, num_dim, -1], [input_shape[2]]]))#-1 的应用:-1 表示不知道该填什么数字合适的情况下
# then swap the channel back
to_tf = tf.transpose(reshaped, [0, 2, 3, 1])#最终转换为这个形式
return to_tf
def _softmax_layer(self, bottom, name):#
if name.startswith('rpn_cls_prob_reshape'):
input_shape = tf.shape(bottom)
bottom_reshaped = tf.reshape(bottom, [-1, input_shape[-1]])
reshaped_score = tf.nn.softmax(bottom_reshaped, name=name)
return tf.reshape(reshaped_score, input_shape)
return tf.nn.softmax(bottom, name=name)
def _proposal_top_layer(self, rpn_cls_prob, rpn_bbox_pred, name):#产生筛选后的roi,不是座标，而是crop出来的，组装成训练数据
with tf.variable_scope(name) as scope:
if cfg.USE_E2E_TF:
rois, rpn_scores = proposal_top_layer_tf(
rpn_cls_prob,
rpn_bbox_pred,
self._im_info,
self._feat_stride,
self._anchors,
self._num_anchors
)
else:#tf.py_func()是一个很重要的扩展tf灵活性的函数，将python函数导入计算图中。参数是（函数，输入，输出，函数是否与状态有关）
#所以其实是调用函数的函数，不同的是，将这个函数导入了计算图中
#func函数中，可以对转化成numpy array的tensor进行np.运算，这就大大扩展了程序的灵活性。对tensor进行np运算。
rois, rpn_scores = tf.py_func(proposal_top_layer,
[rpn_cls_prob, rpn_bbox_pred, self._im_info,
self._feat_stride, self._anchors, self._num_anchors],
[tf.float32, tf.float32], name="proposal_top")
rois.set_shape([cfg.TEST.RPN_TOP_N, 5])
rpn_scores.set_shape([cfg.TEST.RPN_TOP_N, 1])
return rois, rpn_scores
def _proposal_layer(self, rpn_cls_prob, rpn_bbox_pred, name):#proposal_layer_tf输出proposal及其得分，就是rpn计算的roi区域
with tf.variable_scope(name) as scope:
if cfg.USE_E2E_TF:
rois, rpn_scores = proposal_layer_tf(
rpn_cls_prob,
rpn_bbox_pred,
self._im_info,
self._mode,
self._feat_stride,
self._anchors,
self._num_anchors
)
else:
rois, rpn_scores = tf.py_func(proposal_layer,
[rpn_cls_prob, rpn_bbox_pred, self._im_info, self._mode,
self._feat_stride, self._anchors, self._num_anchors],
[tf.float32, tf.float32], name="proposal")
rois.set_shape([None, 5])
rpn_scores.set_shape([None, 1])
return rois, rpn_scores
# Only use it if you have roi_pooling op written in tf.image
def _roi_pool_layer(self, bootom, rois, name):#roi pooling
with tf.variable_scope(name) as scope:
return tf.image.roi_pooling(bootom, rois,
pooled_height=cfg.POOLING_SIZE,
pooled_width=cfg.POOLING_SIZE,
spatial_scale=1. / 16.)[0]
def _crop_pool_layer(self, bottom, rois, name):#
with tf.variable_scope(name) as scope:
batch_ids = tf.squeeze(tf.slice(rois, [0, 0], [-1, 1], name="batch_id"), [1])
# Get the normalized coordinates of bounding boxes
bottom_shape = tf.shape(bottom)
height = (tf.to_float(bottom_shape[1]) - 1.) * np.float32(self._feat_stride[0])
width = (tf.to_float(bottom_shape[2]) - 1.) * np.float32(self._feat_stride[0])
x1 = tf.slice(rois, [0, 1], [-1, 1], name="x1") / width
y1 = tf.slice(rois, [0, 2], [-1, 1], name="y1") / height
x2 = tf.slice(rois, [0, 3], [-1, 1], name="x2") / width
y2 = tf.slice(rois, [0, 4], [-1, 1], name="y2") / height
# Won't be back-propagated to rois anyway, but to save time
bboxes = tf.stop_gradient(tf.concat([y1, x1, y2, x2], axis=1))
pre_pool_size = cfg.POOLING_SIZE * 2
crops = tf.image.crop_and_resize(bottom, bboxes, tf.to_int32(batch_ids), [pre_pool_size, pre_pool_size], name="crops")
return slim.max_pool2d(crops, [2, 2], padding='SAME')
def _dropout_layer(self, bottom, name, ratio=0.5):#dropout操作
return tf.nn.dropout(bottom, ratio, name=name)
def _anchor_target_layer(self, rpn_cls_score, name):# 对rpn的输出进行处理，打上标签
with tf.variable_scope(name) as scope:
rpn_labels, rpn_bbox_targets, rpn_bbox_inside_weights, rpn_bbox_outside_weights = tf.py_func(
anchor_target_layer,
[rpn_cls_score, self._gt_boxes, self._im_info, self._feat_stride, self._anchors, self._num_anchors],
[tf.float32, tf.float32, tf.float32, tf.float32],
name="anchor_target")
rpn_labels.set_shape([1, 1, None, None])
rpn_bbox_targets.set_shape([1, None, None, self._num_anchors * 4])
rpn_bbox_inside_weights.set_shape([1, None, None, self._num_anchors * 4])
rpn_bbox_outside_weights.set_shape([1, None, None, self._num_anchors * 4])
rpn_labels = tf.to_int32(rpn_labels, name="to_int32")
self._anchor_targets['rpn_labels'] = rpn_labels
self._anchor_targets['rpn_bbox_targets'] = rpn_bbox_targets
self._anchor_targets['rpn_bbox_inside_weights'] = rpn_bbox_inside_weights
self._anchor_targets['rpn_bbox_outside_weights'] = rpn_bbox_outside_weights
self._score_summaries.update(self._anchor_targets)
return rpn_labels
def _proposal_target_layer(self, rois, roi_scores, name):#为roi打上具体类别标签
with tf.variable_scope(name) as scope:
rois, roi_scores, labels, bbox_targets, bbox_inside_weights, bbox_outside_weights = tf.py_func(
proposal_target_layer,
[rois, roi_scores, self._gt_boxes, self._num_classes],
[tf.float32, tf.float32, tf.float32, tf.float32, tf.float32, tf.float32],
name="proposal_target")
rois.set_shape([cfg.TRAIN.BATCH_SIZE, 5])
roi_scores.set_shape([cfg.TRAIN.BATCH_SIZE])
labels.set_shape([cfg.TRAIN.BATCH_SIZE, 1])
bbox_targets.set_shape([cfg.TRAIN.BATCH_SIZE, self._num_classes * 4])
bbox_inside_weights.set_shape([cfg.TRAIN.BATCH_SIZE, self._num_classes * 4])
bbox_outside_weights.set_shape([cfg.TRAIN.BATCH_SIZE, self._num_classes * 4])
self._proposal_targets['rois'] = rois
self._proposal_targets['labels'] = tf.to_int32(labels, name="to_int32")
self._proposal_targets['bbox_targets'] = bbox_targets
self._proposal_targets['bbox_inside_weights'] = bbox_inside_weights
self._proposal_targets['bbox_outside_weights'] = bbox_outside_weights
self._score_summaries.update(self._proposal_targets)
return rois, roi_scores
def _anchor_component(self):
with tf.variable_scope('ANCHOR_' + self._tag) as scope:
# just to get the shape right
#将原图缩小到特征图的尺寸，向上取整。
#生成的anchor仅仅是一系列大小，现在将它加上偏移量，成为真正的anchor
#https://blog.csdn.net/zziahgf/article/details/79818141 这篇博文里面有关于这一段的解析，很清楚
height = tf.to_int32(tf.ceil(self._im_info[0] / np.float32(self._feat_stride[0])))
width = tf.to_int32(tf.ceil(self._im_info[1] / np.float32(self._feat_stride[0])))
if cfg.USE_E2E_TF:
anchors, anchor_length = generate_anchors_pre_tf(
height,
width,
self._feat_stride,
self._anchor_scales,
self._anchor_ratios
)
else:
anchors, anchor_length = tf.py_func(generate_anchors_pre,
[height, width,
self._feat_stride, self._anchor_scales, self._anchor_ratios],
[tf.float32, tf.int32], name="generate_anchors")
anchors.set_shape([None, 4])
anchor_length.set_shape([])
self._anchors = anchors
self._anchor_length = anchor_length
'''
'''
def _build_network(self, is_training=True):
# select initializers
if cfg.TRAIN.TRUNCATED:#这个参数默认关
initializer = tf.truncated_normal_initializer(mean=0.0, stddev=0.01)#tf.truncated_normal_initializer：均值、方差双限制初始化
initializer_bbox = tf.truncated_normal_initializer(mean=0.0, stddev=0.001)#同上
else:#按方差随机初始化
initializer = tf.random_normal_initializer(mean=0.0, stddev=0.01)#
initializer_bbox = tf.random_normal_initializer(mean=0.0, stddev=0.001)
#改变初始化的方差，会有什么影响？不过用的既然是预训练的模型，那么初始化应该就是预训练的权重了。
net_conv = self._image_to_head(is_training)#定义卷积层，在resnetv1.py里面写好了，这里直接生成卷积计算节点
with tf.variable_scope(self._scope, self._scope):
# build the anchors for the image
self._anchor_component()#卷积之后计算锚点的一系列计算节点
# region proposal network#定义rpn计算结构
rois = self._region_proposal(net_conv, is_training, initializer)
# region of interest pooling#定义roi-pooling计算节点
if cfg.POOLING_MODE == 'crop':
pool5 = self._crop_pool_layer(net_conv, rois, "pool5")#_crop_pool_layer参数为(self, bottom, rois, name):说明crop的，是特征图上的块
else:
raise NotImplementedError
fc7 = self._head_to_tail(pool5, is_training)#定义网络尾部计算结构，从pool5开始
with tf.variable_scope(self._scope, self._scope):#打开变量
# region classification
cls_prob, bbox_pred = self._region_classification(fc7, is_training, #定义回归及分类计算节点
initializer, initializer_bbox)
self._score_summaries.update(self._predictions)#定义预测结果输出节点
return rois, cls_prob, bbox_pred#最终返回结果：位置、类型、偏移量
def _smooth_l1_loss(self, bbox_pred, bbox_targets, bbox_inside_weights, bbox_outside_weights, sigma=1.0, dim=[1]):
#定义smooth_11_loss计算，rpn的回归用的loss函数。
sigma_2 = sigma ** 2#平方
box_diff = bbox_pred - bbox_targets
in_box_diff = bbox_inside_weights * box_diff
abs_in_box_diff = tf.abs(in_box_diff)
smoothL1_sign = tf.stop_gradient(tf.to_float(tf.less(abs_in_box_diff, 1. / sigma_2)))#tf.stop_gradient阻止梯度传播
in_loss_box = tf.pow(in_box_diff, 2) * (sigma_2 / 2.) * smoothL1_sign \
+ (abs_in_box_diff - (0.5 / sigma_2)) * (1. - smoothL1_sign)#计算loss
out_loss_box = bbox_outside_weights * in_loss_box#Loss乘以权重
loss_box = tf.reduce_mean(tf.reduce_sum(
out_loss_box,
axis=dim
))
return loss_box#返回Box的回归loss
def _add_losses(self, sigma_rpn=3.0):#4个loss加和
with tf.variable_scope('LOSS_' + self._tag) as scope:
# RPN, class loss
rpn_cls_score = tf.reshape(self._predictions['rpn_cls_score_reshape'], [-1, 2])
rpn_label = tf.reshape(self._anchor_targets['rpn_labels'], [-1])
rpn_select = tf.where(tf.not_equal(rpn_label, -1))
rpn_cls_score = tf.reshape(tf.gather(rpn_cls_score, rpn_select), [-1, 2])
rpn_label = tf.reshape(tf.gather(rpn_label, rpn_select), [-1])
rpn_cross_entropy = tf.reduce_mean(
tf.nn.sparse_softmax_cross_entropy_with_logits(logits=rpn_cls_score, labels=rpn_label))
# RPN, bbox loss
rpn_bbox_pred = self._predictions['rpn_bbox_pred']
rpn_bbox_targets = self._anchor_targets['rpn_bbox_targets']
rpn_bbox_inside_weights = self._anchor_targets['rpn_bbox_inside_weights']
rpn_bbox_outside_weights = self._anchor_targets['rpn_bbox_outside_weights']
rpn_loss_box = self._smooth_l1_loss(rpn_bbox_pred, rpn_bbox_targets, rpn_bbox_inside_weights,
rpn_bbox_outside_weights, sigma=sigma_rpn, dim=[1, 2, 3])
# RCNN, class loss#
cls_score = self._predictions["cls_score"]
label = tf.reshape(self._proposal_targets["labels"], [-1])
cross_entropy = tf.reduce_mean(tf.nn.sparse_softmax_cross_entropy_with_logits(logits=cls_score, labels=label))
# RCNN, bbox loss
bbox_pred = self._predictions['bbox_pred']
bbox_targets = self._proposal_targets['bbox_targets']
bbox_inside_weights = self._proposal_targets['bbox_inside_weights']
bbox_outside_weights = self._proposal_targets['bbox_outside_weights']
loss_box = self._smooth_l1_loss(bbox_pred, bbox_targets, bbox_inside_weights, bbox_outside_weights)
self._losses['cross_entropy'] = cross_entropy
self._losses['loss_box'] = loss_box
self._losses['rpn_cross_entropy'] = rpn_cross_entropy
self._losses['rpn_loss_box'] = rpn_loss_box
loss = cross_entropy + loss_box + rpn_cross_entropy + rpn_loss_box
regularization_loss = tf.add_n(tf.losses.get_regularization_losses(), 'regu')
self._losses['total_loss'] = loss + regularization_loss
self._event_summaries.update(self._losses)
return loss
def _region_proposal(self, net_conv, is_training, initializer):#rpn的网络部分
rpn = slim.conv2d(net_conv, cfg.RPN_CHANNELS, [3, 3], trainable=is_training, weights_initializer=initializer,
scope="rpn_conv/3x3")#一个3x3卷积，两个1x1卷积
self._act_summaries.append(rpn)
rpn_cls_score = slim.conv2d(rpn, self._num_anchors * 2, [1, 1], trainable=is_training,
weights_initializer=initializer,
padding='VALID', activation_fn=None, scope='rpn_cls_score')
# change it so that the score has 2 as its channel size#分成两个通道，分别计算回归和分类结果。这样看来貌似两个通道的计算节点是没有交叉的。
rpn_cls_score_reshape = self._reshape_layer(rpn_cls_score, 2, 'rpn_cls_score_reshape')
rpn_cls_prob_reshape = self._softmax_layer(rpn_cls_score_reshape, "rpn_cls_prob_reshape")
rpn_cls_pred = tf.argmax(tf.reshape(rpn_cls_score_reshape, [-1, 2]), axis=1, name="rpn_cls_pred")
rpn_cls_prob = self._reshape_layer(rpn_cls_prob_reshape, self._num_anchors * 2, "rpn_cls_prob")
rpn_bbox_pred = slim.conv2d(rpn, self._num_anchors * 4, [1, 1], trainable=is_training,
weights_initializer=initializer,
padding='VALID', activation_fn=None, scope='rpn_bbox_pred')
if is_training:#训练的步骤
rois, roi_scores = self._proposal_layer(rpn_cls_prob, rpn_bbox_pred, "rois")#计算roi
rpn_labels = self._anchor_target_layer(rpn_cls_score, "anchor")#得到rpn标签
# Try to have a deterministic order for the computing graph, for reproducibility
#考虑到计算图的可重用，要有一个确定的顺序，如下：
with tf.control_dependencies([rpn_labels]):
rois, _ = self._proposal_target_layer(rois, roi_scores, "rpn_rois")#首先计算_proposal_target_layer
else:#测试模式：首先nms一直
if cfg.TEST.MODE == 'nms':
rois, _ = self._proposal_layer(rpn_cls_prob, rpn_bbox_pred, "rois")
elif cfg.TEST.MODE == 'top':#然后处理roi
rois, _ = self._proposal_top_layer(rpn_cls_prob, rpn_bbox_pred, "rois")
else:
raise NotImplementedError
#赋值
self._predictions["rpn_cls_score"] = rpn_cls_score
self._predictions["rpn_cls_score_reshape"] = rpn_cls_score_reshape
self._predictions["rpn_cls_prob"] = rpn_cls_prob
self._predictions["rpn_cls_pred"] = rpn_cls_pred
self._predictions["rpn_bbox_pred"] = rpn_bbox_pred
self._predictions["rois"] = rois
return rois#返回rpn的处理结果roi
def _region_classification(self, fc7, is_training, initializer, initializer_bbox):#最后的分类，输入的fc7
cls_score = slim.fully_connected(fc7, self._num_classes,
weights_initializer=initializer,
trainable=is_training,
activation_fn=None, scope='cls_score')
cls_prob = self._softmax_layer(cls_score, "cls_prob")
cls_pred = tf.argmax(cls_score, axis=1, name="cls_pred")
bbox_pred = slim.fully_connected(fc7, self._num_classes * 4, #最后的回归，输入也是fc7
weights_initializer=initializer_bbox,
trainable=is_training,
activation_fn=None, scope='bbox_pred')
self._predictions["cls_score"] = cls_score
self._predictions["cls_pred"] = cls_pred
self._predictions["cls_prob"] = cls_prob
self._predictions["bbox_pred"] = bbox_pred
return cls_prob, bbox_pred#输出最终分类回归结果。
#下面两个是空的，为了好看，其实没有封装image_to_head和head_to_tail
def _image_to_head(self, is_training, reuse=None):
raise NotImplementedError
def _head_to_tail(self, pool5, is_training, reuse=None):
raise NotImplementedError
def create_architecture(self, mode, num_classes, tag=None, #输出层的结果。用了tf.placeholder机制
anchor_scales=(8, 16, 32), anchor_ratios=(0.5, 1, 2)):
self._image = tf.placeholder(tf.float32, shape=[1, None, None, 3])
self._im_info = tf.placeholder(tf.float32, shape=[3])
self._gt_boxes = tf.placeholder(tf.float32, shape=[None, 5])
self._tag = tag
self._num_classes = num_classes
self._mode = mode
self._anchor_scales = anchor_scales
self._num_scales = len(anchor_scales)
self._anchor_ratios = anchor_ratios
self._num_ratios = len(anchor_ratios)
self._num_anchors = self._num_scales * self._num_ratios
training = mode == 'TRAIN'
testing = mode == 'TEST'
assert tag != None
# handle most of the regularizers here
weights_regularizer = tf.contrib.layers.l2_regularizer(cfg.TRAIN.WEIGHT_DECAY)
if cfg.TRAIN.BIAS_DECAY:
biases_regularizer = weights_regularizer
else:
biases_regularizer = tf.no_regularizer
# list as many types of layers as possible, even if they are not used now
with arg_scope([slim.conv2d, slim.conv2d_in_plane, \
slim.conv2d_transpose, slim.separable_conv2d, slim.fully_connected],
weights_regularizer=weights_regularizer,
biases_regularizer=biases_regularizer,
biases_initializer=tf.constant_initializer(0.0)):
rois, cls_prob, bbox_pred = self._build_network(training)
layers_to_output = {'rois': rois}
for var in tf.trainable_variables():
self._train_summaries.append(var)
if testing:
stds = np.tile(np.array(cfg.TRAIN.BBOX_NORMALIZE_STDS), (self._num_classes))
means = np.tile(np.array(cfg.TRAIN.BBOX_NORMALIZE_MEANS), (self._num_classes))
self._predictions["bbox_pred"] *= stds
self._predictions["bbox_pred"] += means
else:
self._add_losses()
layers_to_output.update(self._losses)
val_summaries = []
with tf.device("/cpu:0"):
val_summaries.append(self._add_gt_image_summary())
for key, var in self._event_summaries.items():
val_summaries.append(tf.summary.scalar(key, var))
for key, var in self._score_summaries.items():
self._add_score_summary(key, var)
for var in self._act_summaries:
self._add_act_summary(var)
for var in self._train_summaries:
self._add_train_summary(var)
self._summary_op = tf.summary.merge_all()
self._summary_op_val = tf.summary.merge(val_summaries)
layers_to_output.update(self._predictions)
return layers_to_output
def get_variables_to_restore(self, variables, var_keep_dic):
raise NotImplementedError
def fix_variables(self, sess, pretrained_model):
raise NotImplementedError
# Extract the head feature maps, for example for vgg16 it is conv5_3
# only useful during testing mode
def extract_head(self, sess, image):
feed_dict = {self._image: image}
feat = sess.run(self._layers["head"], feed_dict=feed_dict)#feed_dict=feed_dict指定数据的头尾
return feat#feature map
# only useful during testing mode#这里feed_dict=feed_dict的头尾，链式调用参数
def test_image(self, sess, image, im_info):
feed_dict = {self._image: image,
self._im_info: im_info}
cls_score, cls_prob, bbox_pred, rois = sess.run([self._predictions["cls_score"],
self._predictions['cls_prob'],
self._predictions['bbox_pred'],
self._predictions['rois']],
feed_dict=feed_dict)
return cls_score, cls_prob, bbox_pred, rois#测试的时候运行这个就ok，这个函数就是执行图运算。被执行的有上面的运算节点及支持其的节点
def get_summary(self, sess, blobs):#这个没明白，summary指的是什么 print总结吗
feed_dict = {self._image: blobs['data'], self._im_info: blobs['im_info'],
self._gt_boxes: blobs['gt_boxes']}
summary = sess.run(self._summary_op_val, feed_dict=feed_dict)
return summary
def train_step(self, sess, blobs, train_op):#训练执行这个
feed_dict = {self._image: blobs['data'], self._im_info: blobs['im_info'],
self._gt_boxes: blobs['gt_boxes']}
rpn_loss_cls, rpn_loss_box, loss_cls, loss_box, loss, _ = sess.run([self._losses["rpn_cross_entropy"],
self._losses['rpn_loss_box'],
self._losses['cross_entropy'],
self._losses['loss_box'],
self._losses['total_loss'],
train_op],
feed_dict=feed_dict)
return rpn_loss_cls, rpn_loss_box, loss_cls, loss_box, loss
def train_step_with_summary(self, sess, blobs, train_op):#孙连的简报？
feed_dict = {self._image: blobs['data'], self._im_info: blobs['im_info'],
self._gt_boxes: blobs['gt_boxes']}
rpn_loss_cls, rpn_loss_box, loss_cls, loss_box, loss, summary, _ = sess.run([self._losses["rpn_cross_entropy"],
self._losses['rpn_loss_box'],
self._losses['cross_entropy'],
self._losses['loss_box'],
self._losses['total_loss'],
self._summary_op,
train_op],
feed_dict=feed_dict)
return rpn_loss_cls, rpn_loss_box, loss_cls, loss_box, loss, summary
def train_step_no_return(self, sess, blobs, train_op):
feed_dict = {self._image: blobs['data'], self._im_info: blobs['im_info'],
self._gt_boxes: blobs['gt_boxes']}
sess.run([train_op], feed_dict=feed_dict)