这里的我就基本看不懂了，除了一些调用其他的函数之外。
主要的是因为对于tensorflow的操作生疏，很多函数的用法都不知道，在更熟练的使用tensorflow后，在回头来看看
感谢WYX同志，该代码注释都是WYX同志完成的
# --------------------------------------------------------
# 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
class Network(object):
def __init__(self):
self._predictions = {} # 保存预测结果
self._losses = {} # 保存损失值
self._anchor_targets = {} # 保存预设anchor的坐标
self._proposal_targets = {}
self._layers = {} # 保存网络
self._gt_image = None #
self._act_summaries = []
self._score_summaries = {}
self._train_summaries = []
self._event_summaries = {}
self._variables_to_fix = {} # 保存fine-tune时需要固定值的变量
# 还原图像，加上均值并进行通道改变
def _add_gt_image(self):
# add back mean
# 预处理时去均值，此处重新加回均值
image = self._image + cfg.PIXEL_MEANS
# BGR to RGB (opencv uses BGR)
# 预处理时缩放图片，此处放大回去 im_info=[缩放后h，缩放后w，缩放比例]
resized = tf.image.resize_bilinear(image, tf.to_int32(self._im_info[:2] / self._im_info[2]))
# 预处理时RGB转成BGR，此处转回RGB
self._gt_image = tf.reverse(resized, axis=[-1])
def _add_gt_image_summary(self):
# use a customized visualization function to visualize the boxes
if self._gt_image is None:
self._add_gt_image()
# image是绘制了box和类别文字后的图片
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):
tf.summary.histogram('ACT/' + tensor.op.name + '/activations', tensor)
# 作用是将输入的Tensor中0元素在所有元素中所占的比例计算并返回
# 因为relu激活函数有时会大面积的将输入参数设为0，所以此函数可以有效衡量relu激活函数的有效性。
tf.summary.scalar('ACT/' + tensor.op.name + '/zero_fraction',
tf.nn.zero_fraction(tensor))
def _add_score_summary(self, key, tensor):
tf.summary.histogram('SCORE/' + tensor.op.name + '/' + key + '/scores', tensor)
def _add_train_summary(self, var):
tf.summary.histogram('TRAIN/' + var.op.name, var)
def _reshape_layer(self, bottom, num_dim, name):
input_shape = tf.shape(bottom)
with tf.variable_scope(name) as scope:
# change the channel to the caffe format
# 输入bottom为[b,h,w,2A]
to_caffe = tf.transpose(bottom, [0, 3, 1, 2])
# to_caffe为[b,2A,h,w]
# then force it to have channel 2
# reshaped为[1,2,b*A*h,w]，且[0,0,b*A*h,w]对应bg，[0,1,b*A*h,w]对应fg
reshaped = tf.reshape(to_caffe,
tf.concat(axis=0, values=[[1, num_dim, -1], [input_shape[2]]]))
# then swap the channel back
# to_tf为[1,b*A*h,w,2]
to_tf = tf.transpose(reshaped, [0, 2, 3, 1])
return to_tf
def _softmax_layer(self, bottom, name):
if name.startswith('rpn_cls_prob_reshape'):
# 以上面to_tf为例，to_tf为[b,A*h,w,2]
input_shape = tf.shape(bottom)
# bottom_reshaped为[b*A*h*w,2] 所有像素点的9个anchor全部平铺成列，平铺时遍历的顺序依次是w,h,A,b
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)
# 对rpn计算结果roiproposals的优选
# 当TEST.MODE = 'top'使用proposal_top_layer，
def _proposal_top_layer(self, rpn_cls_prob, rpn_bbox_pred, name):
'''
对rpn计算结果roi proposals的优选
当TEST.MODE = 'top'使用proposal_top_layer
当TEST.MODE = 'nms'使用proposal_layer
默认使用nms，作者说top模式效果更好，但速度慢
'''
with tf.variable_scope(name) as scope:
if cfg.USE_E2E_TF: # 使用tensorflow端到端实现，不用numpy
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:
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
# 当TEST.MODE = 'nms'使用proposal_layer，
def _proposal_layer(self, rpn_cls_prob, rpn_bbox_pred, name):
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
# 默认用_crop_pool_layer实现ROI Pooling
def _roi_pool_layer(self, bootom, rois, name):
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]
# roi_pooling
def _crop_pool_layer(self, bottom, rois, name):
with tf.variable_scope(name) as scope:
# 提取batch_id,并转为一行，为全0，目的是指定所有crop均来自同一张图片（输入本来也就只有一张特征图）
# rois为[batch_ids,xmin,ymin,xmax,ymax]
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)
# self._feat_stride[0]=16 经过4次pooling
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))
# 为了进行max_pooling，将范围扩大到14*14，这样经过下面的max_pooling得出的结果就是7*7
pre_pool_size = cfg.POOLING_SIZE * 2 # __C.POOLING_SIZE = 7
# crops为[num_boxes, crop_height, crop_width, depth]，一个bottom会输出num_boxes个图像
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')
# dropout，概率为ratio
def _dropout_layer(self, bottom, name, ratio=0.5):
return tf.nn.dropout(bottom, ratio, name=name)
# 根据gt给所有预设anchor计算标签和偏移量
def _anchor_target_layer(self, rpn_cls_score, name):
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 # rpn_labels=[1, 1, A * height, width] 整张图所有预设anchor的标签
self._anchor_targets[
'rpn_bbox_targets'] = rpn_bbox_targets # rpn_bbox_targets=[1, height, width, A * 4] 整张图所有anchor的偏移量
self._anchor_targets[
'rpn_bbox_inside_weights'] = rpn_bbox_inside_weights # rpn_bbox_inside_weights=[1, height, width, A * 4] 在图片范围内的边框权重
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):
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])
'''
rpn_rois=[2000,5] [batch_inds,xmin,ymin,xmax,ymax] batch_inds全0 2000在cfg中调整RPN_POST_NMS_TOP_N
rpn_scores=[2000] clsc层前景的softmax值
rois是对rpn_rois区分前背景后筛选出batch_size个，并且前景在前，背景在后重新排列
roi_scores是对rpn_scores的相同处理
labels是batch_size个区域的标签，前景区域的标签与IOU最大的GT的标签相同，背景的标签为0
bbox_targets=[batch_size，num_class*4] 正确类别的坐标为回归值，其余类别的坐标为0
'''
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
# 生成anchors
def _anchor_component(self):
# 生成每张图的anchor
with tf.variable_scope('ANCHOR_' + self._tag) as scope:
# just to get the shape right
# 卷积特征图的尺寸 self._feat_stride[0]=16 经过4次pooling
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])))
# 默认使用tensorflow端到端，不使用numpy
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: # __C.TRAIN.TRUNCATED = False
# 均值为0，标准差为0.01的截断正态分布，如果生成的值大于平均值2个标准偏差的值则丢弃重新选择。
initializer = tf.truncated_normal_initializer(mean=0.0, stddev=0.01)
initializer_bbox = tf.truncated_normal_initializer(mean=0.0, stddev=0.001)
else: # 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)
# 经过特征提取网络，初步提取特征
net_conv = self._image_to_head(is_training) # _image_to_head在resnet_v1.py实现
with tf.variable_scope(self._scope, self._scope):
# build the anchors for the image
# 生成每张图所有的预设anchor坐标
self._anchor_component()
# region proposal network
# RPN层，输出的是proposal_layer筛选后得分比较高的区域坐标
# rois为[batch_ids,xmin,ymin,xmax,ymax]
rois = self._region_proposal(net_conv, is_training, initializer)
# region of interest pooling
# 进行ROI Pooling，输出与rois数量相同个数的7*7特征图（相当于拷贝了len（rois）个net_conv,每个都只保留判断为前景的某一部分）
if cfg.POOLING_MODE == 'crop':
pool5 = self._crop_pool_layer(net_conv, rois, "pool5")
else:
raise NotImplementedError
fc7 = self._head_to_tail(pool5, is_training)
with tf.variable_scope(self._scope, self._scope):
# region classification
cls_prob, bbox_pred = self._region_classification(fc7, is_training,
initializer, initializer_bbox)
self._score_summaries.update(self._predictions)
# rois为[batch_ids,xmin,ymin,xmax,ymax] batch_size个RPN候选区域坐标
# cls_prob：全连接分类softmax
# bbox_pred：全连接回归
return rois, cls_prob, bbox_pred
# 对于回归的loss计算
def _smooth_l1_loss(self, bbox_pred, bbox_targets, bbox_inside_weights, bbox_outside_weights, sigma=1.0, dim=[1]):
# 以rpn层为例：
# bbox_pred [b,h,w,A*4] reg层特征图
# bbox_targets [1, height, width, A * 4] 整张图所有anchor的偏移量
# sigma=3.0
# dim=[1, 2, 3]
sigma_2 = sigma ** 2
box_diff = bbox_pred - bbox_targets
in_box_diff = bbox_inside_weights * box_diff
abs_in_box_diff = tf.abs(in_box_diff)
# tf.less(a,b) a<b返回真，否则返回假
smoothL1_sign = tf.stop_gradient(tf.to_float(tf.less(abs_in_box_diff, 1. / sigma_2)))
# smoothL1_sign用于实现分段函数
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)
out_loss_box = bbox_outside_weights * in_loss_box
# 求和并降为1维
loss_box = tf.reduce_mean(tf.reduce_sum(
out_loss_box,
axis=dim
))
return loss_box
def _add_losses(self, sigma_rpn=3.0):
with tf.variable_scope('LOSS_' + self._tag) as scope:
# RPN, class loss
# [1,b*A*h,w,2] cls层特征图的reshape
rpn_cls_score = tf.reshape(self._predictions['rpn_cls_score_reshape'], [-1, 2])
# rpn_labels=[1, 1, A * height, width] 整张图所有预设anchor的标签 0为负样本 1为正样本 -1为无效样本
rpn_label = tf.reshape(self._anchor_targets['rpn_labels'], [-1])
# 取出有效样本
rpn_select = tf.where(tf.not_equal(rpn_label, -1))
# 获得有效样本的预测得分
rpn_cls_score = tf.reshape(tf.gather(rpn_cls_score, rpn_select), [-1, 2])
# 获得有效样本的标签
rpn_label = tf.reshape(tf.gather(rpn_label, rpn_select), [-1])
# 计算交叉熵损失 RPN_BATCH_SIZE个anchor的均值
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=[b,h,w,A*4] reg层特征图
rpn_bbox_pred = self._predictions['rpn_bbox_pred']
# rpn_bbox_targets=[1, height, width, A * 4] 整张图所有anchor的偏移量
rpn_bbox_targets = self._anchor_targets['rpn_bbox_targets']
# 权重rpn_bbox_inside_weights=__C.TRAIN.RPN_BBOX_INSIDE_WEIGHTS = (1.0, 1.0, 1.0, 1.0)
rpn_bbox_inside_weights = self._anchor_targets['rpn_bbox_inside_weights']
rpn_bbox_outside_weights = self._anchor_targets['rpn_bbox_outside_weights']
# inside_weights全是1，没起作用，outside_weights为有效样本数量的倒数，起到在一个batch内取平均的作用
# rpn每次只处理一张图片，即一个batch，在一张图片上又取了RPN_BATCH_SIZE个anchor，作用就是对这些anchor取平均
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
# RCNN部分，也就是RPN层之后的部分的分类损失
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
# 边框回归损失
bbox_pred = self._predictions['bbox_pred']
bbox_targets = self._proposal_targets['bbox_targets']
# bbox_inside_weights = __C.TRAIN.BBOX_INSIDE_WEIGHTS = (1.0, 1.0, 1.0, 1.0)
# bbox_outside_weights = np.array(bbox_inside_weights > 0).astype(np.float32)
# inside_weights，outside_weights都是1，没起到作用
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
# 获取正则化损失
regularization_loss = tf.add_n(tf.losses.get_regularization_losses(), 'regu')
# 总损失
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层 紧接在特征提取层之后 256通道 权重初始化等设置和特征提取网络是分开处理的，不共享统一参数空间 没有其他默认参数
'''b=1,根据程序推测RPN层每次只能处理一张图片，yml里IMS_PER_BATCH也确实设置为1'''
rpn = slim.conv2d(net_conv, cfg.RPN_CHANNELS, [3, 3], trainable=is_training, weights_initializer=initializer,
scope="rpn_conv/3x3")
# 添加直方图
self._act_summaries.append(rpn)
# 默认self._num_anchors = 9
# rpn_cls_score=[b,h,w,A*2]
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
# rpn_cls_score_reshape=[1,b*A*h,w,2]
rpn_cls_score_reshape = self._reshape_layer(rpn_cls_score, 2, 'rpn_cls_score_reshape')
# rpn_cls_prob_reshape=[1,b*A*h,w,2]
rpn_cls_prob_reshape = self._softmax_layer(rpn_cls_score_reshape, "rpn_cls_prob_reshape")
# rpn_cls_pred=[b*A*h*w,1] 最大得分的序号，即对应的类别
rpn_cls_pred = tf.argmax(tf.reshape(rpn_cls_score_reshape, [-1, 2]), axis=1, name="rpn_cls_pred")
# rpn_cls_prob=[1,b*h,w,2*A]
rpn_cls_prob = self._reshape_layer(rpn_cls_prob_reshape, self._num_anchors * 2, "rpn_cls_prob") # 概率分布
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') # 预测结果
if is_training:
# 将anchor变成proposals，然后进行NMS，并根据在NMS前后根据pre_nms_topN和post_nms_topN筛选出rpn_cls_prob较高的框
# 推测：rpn_cls_prop=[batch,h,w,9*2] rpn_bbox_pred=[batch,h,w,9*4]
# 推测：rois=[batch_inds,proposal_xmin,proposal_ymin,proposal_xmax,proposal_ymax]
# 推测：roi_scores=[scores]
# rois=[2000,5] [batch_inds,xmin,ymin,xmax,ymax] batch_inds全0 2000在cfg中调整RPN_POST_NMS_TOP_N
# roi_scores=[2000] clsc层前景的softmax值
rois, roi_scores = self._proposal_layer(rpn_cls_prob, rpn_bbox_pred, "rois")
# rpn_labels=[1,1,A*h,w] anchor的标签 0,1,-1
rpn_labels = self._anchor_target_layer(rpn_cls_score, "anchor")
# Try to have a deterministic order for the computing graph, for reproducibility
with tf.control_dependencies([rpn_labels]):
rois, _ = self._proposal_target_layer(rois, roi_scores, "rpn_rois")
else:
if cfg.TEST.MODE == 'nms':
rois, _ = self._proposal_layer(rpn_cls_prob, rpn_bbox_pred, "rois")
elif cfg.TEST.MODE == 'top':
rois, _ = self._proposal_top_layer(rpn_cls_prob, rpn_bbox_pred, "rois")
else:
raise NotImplementedError
# 计算得到的中间结果保存到self._predictions里面，在计算loss的时候或者测试的时候都会用到
self._predictions["rpn_cls_score"] = rpn_cls_score # [b,h,w,A*2] cls层特征图
self._predictions["rpn_cls_score_reshape"] = rpn_cls_score_reshape # [1,b*A*h,w,2] cls层特征图的reshape
self._predictions["rpn_cls_prob"] = rpn_cls_prob # [1,b*h,w,2*A] cls层特征图的softmax
self._predictions["rpn_cls_pred"] = rpn_cls_pred # [b*A*h*w,1] 类别序号
self._predictions["rpn_bbox_pred"] = rpn_bbox_pred # [b,h,w,A*4] reg层特征图
self._predictions["rois"] = rois # batch_size个RPN输出的候选区域坐标
return rois
# 最后的回归和分类层
def _region_classification(self, fc7, is_training, initializer, initializer_bbox):
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,
weights_initializer=initializer_bbox,
trainable=is_training,
activation_fn=None, scope='bbox_pred')
self._predictions["cls_score"] = cls_score # 全连接分类输出
self._predictions["cls_pred"] = cls_pred # 全连接预测的类别序号
self._predictions["cls_prob"] = cls_prob # 全连接softmax
self._predictions["bbox_pred"] = bbox_pred # 全连接回归输出
return cls_prob, bbox_pred
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,
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 # 类别数
self._mode = mode # 模式，nms或top
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
# 正则化参数
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
# 获得模型的输出
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:
# np.title 将矩阵横向复制self._num_classes次
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))
# 对框进行修正
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"):
# 'GROUND_TRUTH'目录下添加标注后的图片
val_summaries.append(self._add_gt_image_summary())
# _event_summaries 包含 self._losses
for key, var in self._event_summaries.items():
# 添加标量
val_summaries.append(tf.summary.scalar(key, var))
# self._score_summaries 包含self._anchor_targets、self._proposal_targets、self._predictions
for key, var in self._score_summaries.items():
# 'SCORE/'目录下添加直方图
self._add_score_summary(key, var)
# self._act_summaries rpn_conv/3x3的输出特征图
for var in self._act_summaries:
# 'ACT/'目录下添加直方图
self._add_act_summary(var)
# self._train_summaries 包含所有可训练变量
for var in self._train_summaries:
# 'TRAIN/'目录下添加直方图
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)
# 模型输出，包含'rois'，self._losses，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)
return feat
# only useful during testing mode
# 输入是仅包含单张图片的blob，用于测试网络
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
# 计算summary 训练过程中进行验证时使用
def get_summary(self, sess, blobs):
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
# 不包含summary的训练op 正常训练时使用
def train_step(self, sess, blobs, train_op):
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
# 计算summary的训练op 满足保存summary的间隔，需要保存summary时使用
def train_step_with_summary(self, sess, blobs, train_op):
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
# 无返回值的训练op
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)
3524
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
