tensorflow学习笔记（四）
代码学习有点吃力，学习了YOLOv1的代码，主要是训练部分的代码，对yolo的又有了进一步的理解。其文件夹下主要包含py文件为，train.py, yolo_net.py, pascal_voc.。下面是比较详细的代码解读。但是还是有一些内容理解的不是很透彻。暂时就这样吧。
首先看一下yolo_net.py文件，这个文件主要定义了网络结构，损失函数的计算等内容。
import numpy as np
import tensorflow as tf
import tensorflow.contrib.slim as slim
'''YOLOnet主要包含了损失函数的计算 和 定义网络结构，其中损失函数的计算包括了交并比函数'''
class YOLONet(object):
def __init__(self, is_training=True):
self.classes = ['aeroplane', 'bicycle', 'bird', 'boat', 'bottle', 'bus', 'car', 'cat', 'chair', 'cow', 'diningtable', 'dog', 'horse',
'motorbike', 'person', 'pottedplant', 'sheep', 'sofa', 'train', 'tvmonitor']
#self.num_class = len(self.classes)
self.num_class = 20
self.image_size = 448 #图片尺寸
self.cell_size = 7 #格子数目
self.boxes_per_cell = 2 #每个格子预测2个框
self.output_size = (self.cell_size * self.cell_size) * (self.num_class + self.boxes_per_cell * 5) #输出的维度S*S*(B*5+C) = 1470
self.scale = 1.0 * self.image_size / self.cell_size #每个格子的像素大小
self.boundary1 = self.cell_size * self.cell_size * self.num_class # 7 * 7 * 20
self.boundary2 = self.boundary1 + self.cell_size * self.cell_size * self.boxes_per_cell #7 * 7 * 20 + 7 * 7 *2
self.object_scale = 1 # 这四个是损失函数前面的系数
self.noobject_scale = 1
self.class_scale = 2
self.coord_scale = 5
self.learning_rate = 0.0001
self.batch_size = 20
self.alpha = 0.1
#offset.shape = [7,7,2]
self.offset = np.transpose(np.reshape(np.array([np.arange(self.cell_size)] * self.cell_size * self.boxes_per_cell),
(self.boxes_per_cell, self.cell_size, self.cell_size)), (1, 2, 0))
self.images = tf.placeholder(tf.float32, [None, self.image_size, self.image_size, 3], name='images')
self.logits = self.build_network(self.images, num_outputs=self.output_size, alpha=self.alpha, is_training=is_training) #self.logits.shape: (?, 1470)
print("self.logits.shape :",self.logits.shape)
if is_training:
#self.labels.shape = [None,7,7,25]
self.labels = tf.placeholder(tf.float32, [None, self.cell_size, self.cell_size, 5 + self.num_class])
# self.logits.shape: (?, 1470)
self.loss_layer(self.logits, self.labels)
self.total_loss = tf.losses.get_total_loss()
tf.summary.scalar('total_loss', self.total_loss)
def build_network(self, images, num_outputs, alpha, keep_prob=0.5, is_training=True):
# num_outputs = 1470
with tf.variable_scope('yolo'):
with slim.arg_scope(
[slim.conv2d, slim.fully_connected],
activation_fn=leaky_relu(alpha),
weights_regularizer=slim.l2_regularizer(0.0005),
weights_initializer=tf.truncated_normal_initializer(0.0, 0.01)
):
net = tf. pad(images, np.array([[0, 0], [3, 3], [3, 3], [0, 0]]), name='pad_1')
net = slim. conv2d(net, 64, 7, 2, padding='VALID', scope='conv_2')
net = slim.max_pool2d(net, 2, padding='SAME', scope='pool_3')
net = slim. conv2d(net, 192, 3, scope='conv_4')
net = slim.max_pool2d(net, 2, padding='SAME', scope='pool_5')
net = slim. conv2d(net, 128, 1, scope='conv_6')
net = slim. conv2d(net, 256, 3, scope='conv_7')
net = slim. conv2d(net, 256, 1, scope='conv_8')
net = slim. conv2d(net, 512, 3, scope='conv_9')
net = slim.max_pool2d(net, 2, padding='SAME', scope='pool_10')
net = slim. conv2d(net, 256, 1, scope='conv_11')
net = slim. conv2d(net, 512, 3, scope='conv_12')
net = slim. conv2d(net, 256, 1, scope='conv_13')
net = slim. conv2d(net, 512, 3, scope='conv_14')
net = slim. conv2d(net, 256, 1, scope='conv_15')
net = slim. conv2d(net, 512, 3, scope='conv_16')
net = slim. conv2d(net, 256, 1, scope='conv_17')
net = slim. conv2d(net, 512, 3, scope='conv_18')
net = slim. conv2d(net, 512, 1, scope='conv_19')
net = slim. conv2d(net, 1024, 3, scope='conv_20')
net = slim.max_pool2d(net, 2, padding='SAME', scope='pool_21')
net = slim. conv2d(net, 512, 1, scope='conv_22')
net = slim. conv2d(net, 1024, 3, scope='conv_23')
net = slim. conv2d(net, 512, 1, scope='conv_24')
net = slim. conv2d(net, 1024, 3, scope='conv_25')
print(net.op.name, net.shape)
net = slim. conv2d(net, 1024, 3, scope='conv_26')
print(net.op.name, net.shape)
net = tf. pad(net, np.array([[0, 0], [1, 1], [1, 1], [0, 0]]), name='pad_27')
print(net.op.name, net.shape)
net = slim . conv2d(net, 1024, 3, 2, padding='VALID', scope='conv_28')
print(net.op.name, net.shape)
net = slim. conv2d(net, 1024, 3, scope='conv_29')
print(net.op.name, net.shape)
net = slim. conv2d(net, 1024, 3, scope='conv_30')
print(net.op.name ,net.shape)
net = tf. transpose(net, [0, 3, 1, 2], name='trans_31')
print(net.op.name, net.shape)
net = slim. flatten(net, scope='flat_32')
print(net.op.name, net.shape)
net = slim.fully_connected(net, 512, scope='fc_33')
print(net.op.name, net.shape)
net = slim.fully_connected(net, 4096, scope='fc_34')
print(net.op.name, net.shape)
net = slim. dropout(net, keep_prob=keep_prob, is_training=is_training, scope='dropout_35')
print(net.op.name, net.shape)
net = slim.fully_connected(net, num_outputs, activation_fn=None, scope='fc_36')
print(net.op.name, net.shape)
return net
def calc_iou(self, boxes1, boxes2, scope='iou'):
"""calculate ious
Args:
boxes1: 5-D tensor [BATCH_SIZE, CELL_SIZE, CELL_SIZE, BOXES_PER_CELL, 4] ====> (x_center, y_center, w, h)
boxes2: 5-D tensor [BATCH_SIZE, CELL_SIZE, CELL_SIZE, BOXES_PER_CELL, 4] ===> (x_center, y_center, w, h)
Return:
iou: 4-D tensor [BATCH_SIZE, CELL_SIZE, CELL_SIZE, BOXES_PER_CELL]
"""
with tf.variable_scope(scope):
# transform (x_center, y_center, w, h) to (x1, y1, x2, y2)
boxes1_t = tf.stack([boxes1[..., 0] - boxes1[..., 2] / 2.0,#四个坐标的计算方法：中心点减去二分之一个宽，得到左边坐标 x1
boxes1[..., 1] - boxes1[..., 3] / 2.0,#中心点减去二分之一个高，得到上坐标 y1
boxes1[..., 0] + boxes1[..., 2] / 2.0,#中心点加上二分之一个高，得到上坐标 y2
boxes1[..., 1] + boxes1[..., 3] / 2.0],#中心点加上二分之一个高，得到上坐标y2
axis=-1)
boxes2_t = tf.stack([boxes2[..., 0] - boxes2[..., 2] / 2.0, #那么下面几行就是计算第二个框的四个坐标值了
boxes2[..., 1] - boxes2[..., 3] / 2.0,
boxes2[..., 0] + boxes2[..., 2] / 2.0,
boxes2[..., 1] + boxes2[..., 3] / 2.0],
axis=-1)
# 计算左上点和右下点
lu = tf.maximum(boxes1_t[..., :2], boxes2_t[..., :2])
rd = tf.minimum(boxes1_t[..., 2:], boxes2_t[..., 2:])
# 计算相交部分面积（我没太弄明白，这个函数到底是怎么计算相交面积的）
intersection = tf.maximum(0.0, rd - lu)
inter_square = intersection[..., 0] * intersection[..., 1]
# 分别计算两个框的面积（真实框和预测框）
square1 = boxes1[..., 2] * boxes1[..., 3]
square2 = boxes2[..., 2] * boxes2[..., 3]
# 这一步在计算两个框相交的面积，公共面积union_square
union_square = tf.maximum(square1 + square2 - inter_square, 1e-10)
# 虽然细节的地方没弄太明白，但是明显该受到这个返回值是交并比，也就是论文中的IOU
return tf.clip_by_value(inter_square / union_square, 0.0, 1.0)
def loss_layer(self, predicts, labels, scope='loss_layer'):
# self.logits.shape : (?, 1470) 预测的是两个框，和20个类别的概率，一共30维
# labels.shape = [None,7,7,25] 真实的图片，只有一个类别和一个框，所以25维
with tf.variable_scope(scope):
# 哦，这里原来就是花式索引那部分：将7*7*30个向量，花式索引
predict_classes = tf.reshape(
predicts[:, :self.boundary1], # boundary1 = 7*7*20
[self.batch_size, self.cell_size, self.cell_size, self.num_class]) # shape=(None,7,7,20)
predict_scales = tf.reshape(
predicts[:, self.boundary1:self.boundary2], # 索引第21和22个，意义是预测的概率大小
[self.batch_size, self.cell_size, self.cell_size, self.boxes_per_cell])
predict_boxes = tf.reshape(predicts[:, self.boundary2:], # 索引框的位置，22之后的元素
[self.batch_size, self.cell_size, self.cell_size, self.boxes_per_cell, 4])#predict_boxes.shape=(None,7,7,2,4)
response = tf.reshape(labels[..., 0], [self.batch_size, self.cell_size, self.cell_size, 1])
# 这些定义的都是框的维度，里面还没有具体的内容
boxes = tf.reshape(labels[..., 1:5], [self.batch_size, self.cell_size, self.cell_size, 1, 4]) # boxes.shape=(None,7,7,1,4)
#那么我这个boxes应该就是真实的框，然后经过了一个归一化
boxes = tf.tile(boxes, [1, 1, 1, self.boxes_per_cell, 1]) / self.image_size # boxes.shape=(None,7,7,2,4)
classes = labels[..., 5:]
# offset.shapae(7,7,2) ----->(1,7,7,2)
offset = tf.reshape(tf.constant(self.offset, dtype=tf.float32), [1, self.cell_size, self.cell_size, self.boxes_per_cell])
# offset.shape(1,7,7,2) ----->(None,7,7,2)
offset = tf.tile(offset, [self.batch_size, 1, 1, 1])
# offset_tran.shape(None,7,7,2)----->(None,7,7,2) 不知道为何7和7要互换一下位置
offset_tran = tf.transpose(offset, (0, 2, 1, 3))
predict_boxes_tran = tf.stack(
[(predict_boxes[..., 0] + offset) / self.cell_size,
(predict_boxes[..., 1] + offset_tran) / self.cell_size,
tf.square(predict_boxes[..., 2]),
tf.square(predict_boxes[..., 3])], axis=-1)
iou_predict_truth = self.calc_iou(predict_boxes_tran, boxes)
# calculate I tensor [BATCH_SIZE, CELL_SIZE, CELL_SIZE, BOXES_PER_CELL]
object_mask = tf.reduce_max(iou_predict_truth, 3, keep_dims=True)
object_mask = tf.cast((iou_predict_truth >= object_mask), tf.float32) * response
# calculate no_I tensor [CELL_SIZE, CELL_SIZE, BOXES_PER_CELL]
noobject_mask = tf.ones_like(object_mask, dtype=tf.float32) - object_mask
# boxes.shape=(None,7,7,2,4)那么最后一个4，是四维的，（x，y，w，h）
boxes_tran = tf.stack(
[boxes[..., 0] * self.cell_size - offset,
boxes[..., 1] * self.cell_size - offset_tran,
tf.sqrt(boxes[..., 2]),
tf.sqrt(boxes[..., 3])], axis=-1)
# class_loss
class_delta = response * (predict_classes - classes)
class_loss = tf.reduce_mean(
tf.reduce_sum(tf.square(class_delta), axis=[1, 2, 3]),
name='class_loss') * self.class_scale
# object_loss
object_delta = object_mask * (predict_scales - iou_predict_truth)
object_loss = tf.reduce_mean(tf.reduce_sum(tf.square(object_delta), axis=[1, 2, 3]),
name='object_loss') * self.object_scale
# noobject_loss
noobject_delta = noobject_mask * predict_scales
noobject_loss = tf.reduce_mean(
tf.reduce_sum(tf.square(noobject_delta), axis=[1, 2, 3]),
name='noobject_loss') * self.noobject_scale
# coord_loss
coord_mask = tf.expand_dims(object_mask, 4)
boxes_delta = coord_mask * (predict_boxes - boxes_tran)
coord_loss = tf.reduce_mean(
tf.reduce_sum(tf.square(boxes_delta), axis=[1, 2, 3, 4]),
name='coord_loss') * self.coord_scale
# 累加loss
tf.losses.add_loss(class_loss)
tf.losses.add_loss(object_loss)
tf.losses.add_loss(noobject_loss)
tf.losses.add_loss(coord_loss)
# summary基本上都是把数据记录到可视化文件的功能
tf.summary.scalar('class_loss', class_loss)
tf.summary.scalar('object_loss', object_loss)
tf.summary.scalar('noobject_loss', noobject_loss)
tf.summary.scalar('coord_loss', coord_loss)
tf.summary.histogram('boxes_delta_x', boxes_delta[..., 0])
tf.summary.histogram('boxes_delta_y', boxes_delta[..., 1])
tf.summary.histogram('boxes_delta_w', boxes_delta[..., 2])
tf.summary.histogram('boxes_delta_h', boxes_delta[..., 3])
tf.summary.histogram('iou', iou_predict_truth)
#重载了leaky_relu函数
def leaky_relu(alpha):
def op(inputs):
return tf.nn.leaky_relu(inputs, alpha=alpha, name='leaky_relu')
return op
第二个文件是pascal_voc.py,这个文件主要定义了一些数据处理的函数，以及从xml文件中提取标签信息的方法。
#将尺寸缩放到448*448并进行归一化到（-1,1） 和对应的labels的数据
import os
import xml.etree.ElementTree as ET # 用于解析xml文件的
import numpy as np
import cv2
import pickle
import copy
class pascal_voc(object): # 定义一个pascal_voc类
def __init__(self, phase, rebuild=False):
self.devkil_path = os.path.join('E:\Desktop\yolo_tensorflow-master1\data\pascal_voc', 'VOCdevkit') # 开发包列表目录：当前工作路径/data/pascal_voc/VOCdevkit
self.data_path = os.path.join(self.devkil_path, 'VOC2007') # 开发包数据目录：当前工作路径/data/pascal_voc/VOCdevkit/VOC2007
self.cache_path = 'E:\Desktop\yolo_tensorflow-master1\data\pascal_voc\cache' # 见yolo目录下的config.py文件
self.batch_size = 20
self.image_size = 448
self.cell_size = 7
self.classes = ['aeroplane', 'bicycle', 'bird', 'boat', 'bottle', 'bus', 'car', 'cat', 'chair', 'cow', 'diningtable', 'dog', 'horse',
'motorbike', 'person', 'pottedplant', 'sheep', 'sofa', 'train', 'tvmonitor']
# 将类别中文名数字序列化成0,1,2,……
self.class_to_ind = dict(zip(self.classes, range(len(self.classes))))
self.flipped = True # 水平翻转的参数，应该是想增强数据
self.phase = phase # 定义训练or测试
self.rebuild = rebuild
self.cursor = 0 # 光标移动用，查询gt_labels这个结构
self.epoch = 1
self.gt_labels = None
self.prepare()
def get(self):
# 初始化图像。batch_size x 448x448x3， self.batch_size=30, image_size=448
images = np.zeros((self.batch_size, self.image_size, self.image_size, 3))
# 初始化类别（gt）。batch_size x 7x7x25 ,cell_size = 7,对于另外一个box就不构建维度了，因此是25
labels = np.zeros((self.batch_size, self.cell_size, self.cell_size, 25))
count = 0
while count < self.batch_size: # batch处理
imname = self.gt_labels[self.cursor]['imname'] # 从gt label中读取图像名
flipped = self.gt_labels[self.cursor]['flipped'] # 从gt label中查看是否flipped
images[count, :, :, :] = self.image_read(imname, flipped)
# 从gt label中获取label类别坐标等信息
labels[count, :, :, :] = self.gt_labels[self.cursor]['label']
count += 1
self.cursor += 1
if self.cursor >= len(self.gt_labels): # 判断是否训练完一个epoch了
np.random.shuffle(self.gt_labels)
self.cursor = 0
self.epoch += 1
return images, labels # 返回尺寸缩放和归一化后的image序列；以及labels 真实信息
#读取图片做归一化
def image_read(self, imgname, flipped=False):
image = cv2.imread(imgname)
image = cv2.resize(image, (self.image_size, self.image_size))
# astype,转换数据类型
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB).astype(np.float32)
# 图像像素值归一化到(-1,1)
image = (image / 255.0) * 2.0 - 1.0
if flipped:
image = image[:, ::-1, :]
return image
def prepare(self): # 是否做flipped并打乱原来次序返回结果
gt_labels = self.load_labels() # 获取gt labels数据
if self.flipped: # 判断是否做flipped
print('Appending horizontally-flipped training examples ...')
gt_labels_cp = copy.deepcopy(gt_labels)
for idx in range(len(gt_labels_cp)):
gt_labels_cp[idx]['flipped'] = True
gt_labels_cp[idx]['label'] = \
gt_labels_cp[idx]['label'][:, ::-1, :]
for i in range(self.cell_size):
for j in range(self.cell_size):
if gt_labels_cp[idx]['label'][i, j, 0] == 1:
gt_labels_cp[idx]['label'][i, j, 1] = \
self.image_size - 1 - \
gt_labels_cp[idx]['label'][i, j, 1]
gt_labels += gt_labels_cp
np.random.shuffle(gt_labels) # 对gt labels打乱顺序
self.gt_labels = gt_labels
return gt_labels
def load_labels(self):
cache_file = os.path.join(
self.cache_path, 'pascal_' + self.phase + '_gt_labels.pkl') # cache/pascal_test/train_gt_labels.pkl
if os.path.isfile(cache_file) and not self.rebuild:
print('Loading gt_labels from: ' + cache_file) # 从cache目录加载gt label文件
with open(cache_file, 'rb') as f:
gt_labels = pickle.load(f)
return gt_labels # 返回gt
print('Processing gt_labels from: ' + self.data_path) # 处理来自data目录下的gt label
if not os.path.exists(self.cache_path): # 如果不存在目录文件则创建
os.makedirs(self.cache_path)
if self.phase == 'train':
# 如果是train阶段，则txtname是：当前工作路径/data/pascal_voc/VOCdevkit/VOC2007/ImageSets/Main/trainval.txt这个
txtname = os.path.join(self.data_path, 'ImageSets', 'Main', 'trainval.txt')
else:
# 如果是test阶段，则txtname是：当前工作路径/data/pascal_voc/VOCdevkit/VOC2007/ImageSets/Main/test.txt这个
txtname = os.path.join(self.data_path, 'ImageSets', 'Main', 'test.txt')
with open(txtname, 'r') as f:
self.image_index = [x.strip() for x in f.readlines()]
gt_labels = [] # 创建列表存放gt label
for index in self.image_index:
print(index)
label, num = self.load_pascal_annotation(index) # 取gt label以及num目标数
if num == 0:
continue
# 找到图像文件夹下对应索引号的图像
imname = os.path.join(self.data_path, 'JPEGImages', index + '.jpg')
gt_labels.append({'imname': imname,
'label': label,
'flipped': False})
print('Saving gt_labels to: ' + cache_file)
with open(cache_file, 'wb') as f:
pickle.dump(gt_labels, f) # 将gt labels（图形名，目标类别位置坐标信息，是否flipped）写入cache中
return gt_labels
def load_pascal_annotation(self, index): # 从xml文件中获取bbox信息
"""
Load image and bounding boxes info from XML file in the PASCAL VOC
format.
"""
imname = os.path.join(self.data_path, 'JPEGImages', index + '.jpg') # 图像目录下读取jpg文件：当前工作路径/data/pascal_voc/VOCdevkit/VOC2007/JPEGImages
im = cv2.imread(imname)
h_ratio = 1.0 * self.image_size / im.shape[0] # 尺寸缩放系数
w_ratio = 1.0 * self.image_size / im.shape[1]
# im = cv2.resize(im, [self.image_size, self.image_size])
label = np.zeros((self.cell_size, self.cell_size, 25))
filename = os.path.join(self.data_path, 'Annotations', index + '.xml') # 读取xml文件
tree = ET.parse(filename) # 解析树
objs = tree.findall('object') # 找xml文件中的object
for obj in objs: # 遍历object
bbox = obj.find('bndbox') # 查找object的bounding box
# Make pixel indexes 0-basedq
# (float(bbox.find('xmin').text) - 1) * w_ratio
x1 = max(min((float(bbox.find('xmin').text) - 1) * w_ratio, self.image_size - 1), 0) # 将xml文件中的坐标做尺寸缩放
y1 = max(min((float(bbox.find('ymin').text) - 1) * h_ratio, self.image_size - 1), 0)
x2 = max(min((float(bbox.find('xmax').text) - 1) * w_ratio, self.image_size - 1), 0)
y2 = max(min((float(bbox.find('ymax').text) - 1) * h_ratio, self.image_size - 1), 0)
cls_ind = self.class_to_ind[obj.find('name').text.lower().strip()] # 实际类别对应数字序号
boxes = [(x2 + x1) / 2.0, (y2 + y1) / 2.0, x2 - x1, y2 - y1] # 坐标转换成(x,y,w,h)
x_ind = int(boxes[0] * self.cell_size / self.image_size) # 判断x属于第几个cell
y_ind = int(boxes[1] * self.cell_size / self.image_size) # 判断y属于第几个cell
if label[y_ind, x_ind, 0] == 1:
print('x_ind{},y_ind{}'.format(x_ind, y_ind))
print('label[y_ind, x_ind, 1:5]{}'.format(label[y_ind, x_ind, 1:5])) # 坐标赋值
print('label[y_ind, x_ind, 0]等于1 ', label[y_ind, x_ind, 0])
continue
print('x_ind{},y_ind{}'.format(x_ind, y_ind))
print('label[y_ind, x_ind, 1:5]{}'.format(label[y_ind, x_ind, 1:5])) # 坐标赋值
print('label[y_ind, x_ind, 0]不等于1 ', label[y_ind, x_ind, 0])
label[y_ind, x_ind, 0] = 1 # cell索引后，是否存在目标位赋1
label[y_ind, x_ind, 1:5] = boxes # 坐标赋值
label[y_ind, x_ind, 5 + cls_ind] = 1 # 类别赋值
return label, len(objs) # 返回label（gt）/以及xml中目标个数第三个是train.py，这个主要定义了一个主函数接口，将数据和网络丢给这个处理机制（类），就可以进行训练了。
import os
import argparse
import datetime
import tensorflow as tf
import yolo.config as cfg
from yolo.yolo_net import YOLONet
from utils.timer import Timer
from utils.pascal_voc import pascal_voc
slim = tf.contrib.slim
class Solver(object):
def __init__(self, net, data):
self.net = net
self.data = data
self.weights_file = 'E:\Desktop\yolo_tensorflow-master1\data\pascal_voc\weights\YOLO_small.ckpt'
self.max_iter = 10000
self.initial_learning_rate = 0.0001
self.decay_steps = 30000
self.decay_rate = 0.1
self.staircase = True
self.summary_iter = 10
self.save_iter = 1000
self.output_dir = os.path.join(
'E:\Desktop\yolo_tensorflow-master1\data\pascal_voc\output', datetime.datetime.now().strftime('%Y_%m_%d_%H_%M'))
if not os.path.exists(self.output_dir):
os.makedirs(self.output_dir)
self.save_cfg()
self.variable_to_restore = tf.global_variables()
self.saver = tf.train.Saver(self.variable_to_restore, max_to_keep=None)
self.ckpt_file = os.path.join(self.output_dir, 'yolo')
self.summary_op = tf.summary.merge_all()
self.writer = tf.summary.FileWriter(self.output_dir, flush_secs=60)
# self.global_step疑似一个计数器，好像没什么太大用处吧。比如说下面的train.op和learning_rate用到了global_step
# 那么作用就是我的train_op每训练一次，或者learning_rate每更新一次，我这个global_step会自动加一，起到一个计数的效果
self.global_step = tf.train.create_global_step()
self.learning_rate = tf.train.exponential_decay(
self.initial_learning_rate, self.global_step, self.decay_steps,
self.decay_rate, self.staircase, name='learning_rate')
self.optimizer = tf.train.GradientDescentOptimizer(learning_rate=self.learning_rate)
self.train_op = slim.learning.create_train_op(self.net.total_loss, self.optimizer, global_step=self.global_step)
gpu_options = tf.GPUOptions()
config = tf.ConfigProto(gpu_options=gpu_options)
self.sess = tf.Session(config=config)
self.sess.run(tf.global_variables_initializer())
if self.weights_file is not None:
print('Restoring weights from: ' + self.weights_file)
self.saver.restore(self.sess, self.weights_file)
self.writer.add_graph(self.sess.graph)
# 定义训练函数
def train(self):
train_timer = Timer() # 计算时间
load_timer = Timer() # 计算时间
for step in range(1, self.max_iter + 1):
load_timer.tic() # 计算时间
images, labels = self.data.get()
load_timer.toc() # 计算时间
feed_dict = {self.net.images: images, self.net.labels: labels}
# 这一层的if是说，每10次我就把我的信息保存一次到模型里，所以else中的内容就是正常的训练过程，当然这个模型是为了进行可视化的吧
# 当然大部分肯定是执行的else部分
if step % self.summary_iter == 0:
# 这一层的if是说，每100次打印出一次结果
if step % (self.summary_iter * 10) == 0:
train_timer.tic() # 计算时间
# 运行模型、损失函数、和train_op
summary_str = self.sess.run(self.summary_op, feed_dict=feed_dict)
loss = self.sess.run(self.net.total_loss, feed_dict=feed_dict)
_ = self.sess.run(self.train_op, feed_dict=feed_dict)
train_timer.toc() # 计算时间
# 打印时间、loss、迭代次数等信息
log_str = '''{} Epoch: {}, Step: {}, Learning rate: {}, Loss: {:5.3f}\nSpeed:
{:.3f}s/iter, Load: {:.3f}s/iter, Remain: {}
'''.format(
datetime.datetime.now().strftime('%m-%d %H:%M:%S'),
self.data.epoch,
int(step),
round(self.learning_rate.eval(session=self.sess), 6),
loss,
train_timer.average_time,
load_timer.average_time,
train_timer.remain(step, self.max_iter))
print(log_str)
else:
train_timer.tic() # 计算时间
summary_str, _ = self.sess.run([self.summary_op, self.train_op], feed_dict=feed_dict)
train_timer.toc() # 计算时间
self.writer.add_summary(summary_str, step)
else:
train_timer.tic() # 计算时间
self.sess.run(self.train_op, feed_dict=feed_dict)
train_timer.toc() # 计算时间
# 每1000次保存一次模型文件，到指定路径
if step % self.save_iter == 0:
print('{} Saving checkpoint file to: {}'.format(
datetime.datetime.now().strftime('%m-%d %H:%M:%S'),
self.output_dir))
self.saver.save(self.sess, self.ckpt_file, global_step=self.global_step)
def save_cfg(self):
with open(os.path.join(self.output_dir, 'config.txt'), 'w') as f:
cfg_dict = cfg.__dict__
for key in sorted(cfg_dict.keys()):
if key[0].isupper():
cfg_str = '{}: {}\n'.format(key, cfg_dict[key])
f.write(cfg_str)
def update_config_paths(data_dir, weights_file):
cfg.DATA_PATH = data_dir
cfg.PASCAL_PATH = os.path.join(data_dir, 'pascal_voc')
cfg.CACHE_PATH = os.path.join('E:\Desktop\yolo_tensorflow-master1\data\pascal_voc', 'cache')
cfg.OUTPUT_DIR = os.path.join('E:\Desktop\yolo_tensorflow-master1\data\pascal_voc', 'output')
cfg.WEIGHTS_DIR = os.path.join('E:\Desktop\yolo_tensorflow-master1\data\pascal_voc', 'weights')
cfg.WEIGHTS_FILE = os.path.join('E:\Desktop\yolo_tensorflow-master1\data\pascal_voc\weights', weights_file)
def main():
'''
虽说这里是main函数，但是实际上他只是定义了一些接口，比如说存放模型文件的路径，是否使用gpu训练等
就是说定义了下面这些东西，你在运行train.py的时候，通过改变下面定义的这些参数，可以更改一些配置
而真正的训练过程，在Solver.train()中
'''
parser = argparse.ArgumentParser() # 添加参数的函数，在终端运行train.py时，输入想要改变的东西
parser.add_argument('--weights', default="YOLO_small.ckpt", type=str)
parser.add_argument('--data_dir', default="data", type=str)
parser.add_argument('--threshold', default=0.2, type=float)
parser.add_argument('--iou_threshold', default=0.5, type=float)
parser.add_argument('--gpu', default='', type=str)
args = parser.parse_args()
# 设置GPU和权重文件路径的
if args.gpu is not None:
cfg.GPU = args.gpu
if args.data_dir != cfg.DATA_PATH:
update_config_paths(args.data_dir, args.weights)
os.environ['CUDA_VISIBLE_DEVICES'] = '0'
# 下面几乎都是各个调用，因为之前定义的都是Class，然后我们要创建实际的对象
yolo = YOLONet() # 首先建立网络模型对象
pascal = pascal_voc('train') # 然后又建立数据集的对象
solver = Solver(yolo, pascal) # 把网络模型和数据集丢给solver
solver.train() # 使用train函数进行训练
# 设置一个主函数借口，可以从外部调用
if __name__ == '__main__':
# python train.py --weights YOLO_small.ckpt --gpu 0
main()
有可能会有少量表述不正当的地方，仅供参考。