From 5f498b3c114e6616aee1cf762d9f85bc6c90e2f5 Mon Sep 17 00:00:00 2001
From: wenlong <wenlong@nfs.iscas.ac.cn>
Date: Mon, 13 Mar 2017 16:20:45 +0800
Subject: [PATCH 1/4] <description> read source code and add some common:
 </description>

---
 .gitignore                     |   1 +
 demo.py                        |  76 +++++++++++++++++++++------------
 tools/preprocess_pascal_voc.py |   2 +
 tools/train.py                 |   2 +
 yolo/__init__.pyc              | Bin 0 -> 203 bytes
 yolo/dataset/__init__.pyc      | Bin 0 -> 195 bytes
 yolo/dataset/dataset.py        |   3 ++
 yolo/dataset/dataset.pyc       | Bin 0 -> 886 bytes
 yolo/dataset/text_dataset.py   |  21 ++++++---
 yolo/dataset/text_dataset.pyc  | Bin 0 -> 4740 bytes
 yolo/net/net.py                |  26 ++++++++---
 yolo/net/yolo_net.py           |  25 ++++++++++-
 yolo/net/yolo_tiny_net.py      |  33 ++++++++++----
 yolo/solver/solver.py          |   3 ++
 yolo/solver/yolo_solver.py     |  21 ++++++---
 yolo/utils/process_config.py   |   6 +++
 16 files changed, 168 insertions(+), 51 deletions(-)
 create mode 100644 .gitignore
 create mode 100644 yolo/__init__.pyc
 create mode 100644 yolo/dataset/__init__.pyc
 create mode 100644 yolo/dataset/dataset.pyc
 create mode 100644 yolo/dataset/text_dataset.pyc

diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000..0d20b64
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1 @@
+*.pyc
diff --git a/demo.py b/demo.py
index bbf8179..3551baa 100644
--- a/demo.py
+++ b/demo.py
@@ -1,3 +1,5 @@
+#!/usr/bin/python 
+# -*- coding: utf-8 -*- 
 import sys
 
 sys.path.append('./')
@@ -7,18 +9,32 @@
 import cv2
 import numpy as np
 
-classes_name =  ["aeroplane", "bicycle", "bird", "boat", "bottle", "bus", "car", "cat", "chair", "cow", "diningtable", "dog", "horse", "motorbike", "person", "pottedplant", "sheep", "sofa", "train","tvmonitor"]
+classes_name =  ["aeroplane", "bicycle", "bird", "boat", "bottle", 
+                "bus", "car", "cat", "chair", "cow", 
+                "diningtable", "dog", "horse", "motorbike", "person", 
+                "pottedplant", "sheep", "sofa", "train","tvmonitor"]
 
+common_params = { 'image_size': 448, 
+                  'num_classes': 20, 
+                  'batch_size':1}
+
+net_params = {'cell_size': 7, 
+              'boxes_per_cell':2, 
+              'weight_decay': 0.0005}
 
 def process_predicts(predicts):
+  """
+  对于规范化的输出结果对于特定的用户可能觉得不习惯，那么实现一个接口，将规范化
+  的结果重新编写为用户习惯的数据类型
+  """
   p_classes = predicts[0, :, :, 0:20]
   C = predicts[0, :, :, 20:22]
   coordinate = predicts[0, :, :, 22:]
-
+  # 训练的模型设置超参数 net_params, 其中cell大小设置为7
   p_classes = np.reshape(p_classes, (7, 7, 1, 20))
   C = np.reshape(C, (7, 7, 2, 1))
 
-  P = C * p_classes
+  P = C * p_classes   # P size = (7, 7, 2, 20)
 
   #print P[5,1, 0, :]
 
@@ -51,36 +67,42 @@ def process_predicts(predicts):
 
   return xmin, ymin, xmax, ymax, class_num
 
-common_params = {'image_size': 448, 'num_classes': 20, 
-                'batch_size':1}
-net_params = {'cell_size': 7, 'boxes_per_cell':2, 'weight_decay': 0.0005}
-
-net = YoloTinyNet(common_params, net_params, test=True)
-
-image = tf.placeholder(tf.float32, (1, 448, 448, 3))
-predicts = net.inference(image)
+def main():
 
-sess = tf.Session()
 
-np_img = cv2.imread('cat.jpg')
-resized_img = cv2.resize(np_img, (448, 448))
-np_img = cv2.cvtColor(resized_img, cv2.COLOR_BGR2RGB)
+  net = YoloTinyNet(common_params, net_params, test=True)
+  # tensorflow中声明占位符号image, 这在后面run的时候
+  # feed_dict中会出现该占位符和对应的值，意思就是输入数据的来源
+  image = tf.placeholder(tf.float32, (1, 448, 448, 3))
+  predicts = net.inference(image)
 
+  sess = tf.Session()
 
-np_img = np_img.astype(np.float32)
+  # 转化数据格式
+  np_img = cv2.imread('cat.jpg')
+  resized_img = cv2.resize(np_img, (448, 448))
+  np_img = cv2.cvtColor(resized_img, cv2.COLOR_BGR2RGB)
 
-np_img = np_img / 255.0 * 2 - 1
-np_img = np.reshape(np_img, (1, 448, 448, 3))
+  np_img = np_img.astype(np.float32)
+  #白化输入的数据
+  np_img = np_img / 255.0 * 2 - 1
+  np_img = np.reshape(np_img, (1, 448, 448, 3))
 
-saver = tf.train.Saver(net.trainable_collection)
+  saver = tf.train.Saver(net.trainable_collection)
 
-saver.restore(sess, 'models/pretrain/yolo_tiny.ckpt')
+  saver.restore(sess, 'models/pretrain/yolo_tiny.ckpt')
+  # The optional feed_dict argument allows the caller to override 
+  # the value of tensors in the graph. 
+  np_predict = sess.run(predicts, feed_dict={image: np_img})
 
-np_predict = sess.run(predicts, feed_dict={image: np_img})
+  xmin, ymin, xmax, ymax, class_num = process_predicts(np_predict)
+  class_name = classes_name[class_num]
+  # 绘制预测框, 输出预测类型
+  cv2.rectangle(resized_img, (int(xmin), int(ymin)), (int(xmax), int(ymax)), (0, 0, 255))
+  cv2.putText(resized_img, 
+              class_name, (int(xmin), int(ymin)), 2, 1.5, (0, 0, 255))
+  cv2.imwrite('cat_out.jpg', resized_img)
+  sess.close()
 
-xmin, ymin, xmax, ymax, class_num = process_predicts(np_predict)
-class_name = classes_name[class_num]
-cv2.rectangle(resized_img, (int(xmin), int(ymin)), (int(xmax), int(ymax)), (0, 0, 255))
-cv2.putText(resized_img, class_name, (int(xmin), int(ymin)), 2, 1.5, (0, 0, 255))
-cv2.imwrite('cat_out.jpg', resized_img)
-sess.close()
+if __name__ == '__main__':
+  main()
\ No newline at end of file
diff --git a/tools/preprocess_pascal_voc.py b/tools/preprocess_pascal_voc.py
index 76d43fc..fee497f 100755
--- a/tools/preprocess_pascal_voc.py
+++ b/tools/preprocess_pascal_voc.py
@@ -1,3 +1,5 @@
+#!/usr/bin/python 
+# -*- coding: utf-8 -*- 
 """preprocess pascal_voc data
 """
 import os
diff --git a/tools/train.py b/tools/train.py
index 5b399ba..572c67b 100644
--- a/tools/train.py
+++ b/tools/train.py
@@ -1,3 +1,5 @@
+#!/usr/bin/python 
+# -*- coding: utf-8 -*- 
 import sys
 from optparse import OptionParser
 
diff --git a/yolo/__init__.pyc b/yolo/__init__.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..6eebee04818f9289749078a326b82d3013973776
GIT binary patch
literal 203
zcmZSn%*!<^;%G!N0~9a;X$K%K)&LSIK*Y$9!@v*)XEQQHF~Y=|7^0X!3W7C2x-$O*
zfu9C5kXgbG(wtb5Se#l?0usmr(riGwI6tQ>wFqQ11A+is<!lublv<D<Q<9ojoL`id
vlV7e|nV*v%1193*GxIV_;^XxSDoZ$kdTny^Q%ZAE?HEBuvH%GlMjl20x63Q_

literal 0
HcmV?d00001

diff --git a/yolo/dataset/__init__.pyc b/yolo/dataset/__init__.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..7dec5bd23b663c719b65ec04b66790e46ea832bb
GIT binary patch
literal 195
zcmZSn%*!<^;%G!N0~9a;X$K%KmH`qeK*Y$9!@v*)XEQQHF@gkxH9#se{{w*^m|4OO
z(vVn^Se#l?!UJTNq*j!~Lj^Q|d<FzjYyjjsTg3#W7Uaj2q~;ap7p3Lom+MyM=j6wL
n38=o9`1s7c%#!$cy@JXT4xlER-29Z%oK!nTAiEgk0v<*H(z7l{

literal 0
HcmV?d00001

diff --git a/yolo/dataset/dataset.py b/yolo/dataset/dataset.py
index de56381..7bd8dd9 100644
--- a/yolo/dataset/dataset.py
+++ b/yolo/dataset/dataset.py
@@ -1,3 +1,6 @@
+#!/usr/bin/python 
+# -*- coding: utf-8 -*- 
+
 """DataSet  base class 
 """
 class DataSet(object):
diff --git a/yolo/dataset/dataset.pyc b/yolo/dataset/dataset.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..14f4f69b06f043d16c5c102cd74142947f1ec49d
GIT binary patch
literal 886
zcmb_a%TB{E5ZokfN+T5xh`W!u0QU+ZXhEo#9@0yRB3WtTilQWr;zg)e;4AnazK9=S
z*A0(KZ~-ED;%vq<v$nn0LGV0(nP<>G9=*?K7(tK&#Doqg2bK=xE<6D|0$j%~<Q~kz
z4l@r3E>cWq3WZptMv1IQjS&I%KL30eI8+<)GC`oofOH8-&|&r1Yi;5VM6;0Nm_33K
z>$Wa|5E=jA!E$37CLln72a<Crnai~a0g;KW%TiaePMfqeLvb!PGm)<{6qb=w14iNZ
z6w8=~?1lRTQ;n0dE>x*1RQYAo=*A{RyC14p;(%CNrnpO$ZIh5HV8I;IY&=XNRqF&*
zWpuMF^!?#l7do-fu1wnZQN6|;S|jDETA`Fa%?-B<-HcPdj~{o2(_GPcEmF*GzO)zd
zO513}a(|52B?;@m+p)>#g<NdIS81uFL_RtxOP${p{K>aQ%3NnsMof;lyU!%Q-m!>e
p?42<z=)1nt?@jHwMSUtDSM}mnWoViI;Aq=C*)(=)ZzO~7yaC*+#QgvO

literal 0
HcmV?d00001

diff --git a/yolo/dataset/text_dataset.py b/yolo/dataset/text_dataset.py
index 15c2b77..d4e886b 100644
--- a/yolo/dataset/text_dataset.py
+++ b/yolo/dataset/text_dataset.py
@@ -1,3 +1,5 @@
+#!/usr/bin/python 
+# -*- coding: utf-8 -*- 
 from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
@@ -12,11 +14,13 @@
 
 from yolo.dataset.dataset import DataSet 
 
+
 class TextDataSet(DataSet):
   """TextDataSet
   process text input file dataset 
   text file format:
     image_path xmin1 ymin1 xmax1 ymax1 class1 xmin2 ymin2 xmax2 ymax2 class2
+  设计思想是采用生产者消费者模式，
   """
 
   def __init__(self, common_params, dataset_params):
@@ -52,11 +56,12 @@ def __init__(self, common_params, dataset_params):
     self.record_number = len(self.record_list)
 
     self.num_batch_per_epoch = int(self.record_number / self.batch_size)
-
+    # 创建生产者守护进程并启动
     t_record_producer = Thread(target=self.record_producer)
     t_record_producer.daemon = True 
     t_record_producer.start()
-
+    
+    # 创建thread_num个消费者守护进程并启动
     for i in range(self.thread_num):
       t = Thread(target=self.record_customer)
       t.daemon = True
@@ -109,6 +114,12 @@ def record_process(self, record):
       labels[object_num] = [xcenter, ycenter, box_w, box_h, class_num]
       object_num += 1
       i += 5
+      # TODO:
+      # 这个地方会不会忽略掉一些显著特征呢？
+      # 因为self.max_objects是自定义的变量，在读取的过程中，
+      # 仅仅读取前面的数据的话，后面的会被忽略掉的。
+      # TODO: 训练数据中每张图片的物体是如何给出的，是根据显著性呢还是根据
+      # 起始点为位置大小给出的呢，这个需要check一下
       if object_num >= self.max_objects:
         break
     return [image, labels, object_num]
@@ -124,9 +135,9 @@ def record_customer(self):
   def batch(self):
     """get batch
     Returns:
-      images: 4-D ndarray [batch_size, height, width, 3]
-      labels: 3-D ndarray [batch_size, max_objects, 5]
-      objects_num: 1-D ndarray [batch_size]
+      images: 4-D ndarray [batch_size, height, width, 3] 一个batch中所有图片数据
+      labels: 3-D ndarray [batch_size, max_objects, 5] 一个batch中的所有图片的中的所有物体的标签
+      objects_num: 1-D ndarray [batch_size] 一个batch中每个图片中object的个数
     """
     images = []
     labels = []
diff --git a/yolo/dataset/text_dataset.pyc b/yolo/dataset/text_dataset.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..7f55d47836bcf79401a8d925707657bed9d1e989
GIT binary patch
literal 4740
zcmb_f-EtdQ5$+lN%l|)d>=-svOBGRSy(wbv0_6=XNlfil0bZDFk*qdmb}WsoaU_jA
za}>*zbrtTZ;*u)fAy>c)aKq(30T(;~RXhOueSJo@6PBALmUYxW=bY)&r~B)k!oTP0
ze|++{CoReU3i$oW7mj3q2GPd9BR!G6BTYvf@~-l(G+oOVlrKoLVELl-iqb6VTF@nF
zmZo%Bn&l~7k!EE|SEX4~x+J~2^y|{B+r7%to00ykG-oYek=~s2=cPFhzADkYbR22d
z<aZ)3L_TXSNV_PH8g-Z+^E?d)gIT~39%e~zWTGJM50lhbMHPy6{5;O$WMC!pkPOp!
zV1mwQ&@zzkTSW<qUyY(sWZ5#<PrGRpwk=-)zZaVDQDn>vb<x=5{Do&}%L97$0}zqW
zq3|c%2p8i3@E?3j0=_FK0ml`UfY(Y&RK$PO;KQ>&y!w>7=MB@O6=j)c@N6$03`fT6
z#J$LC!>%kc9;T>Dg`Fhrhh`Tt&x`xvQ4|bA)Ah#vcyQM{({bDn$D}#7dSRB`RrR*2
zw?T(v(i~O2jm=sNAG)9ZddS1R#IcRYxgK-lg(K%8fg{G1bDZCmsi^SZqT*CkxR_N^
z5%wxxh|1Ghr3+C-OjX)&S_NU@@PW0dq&k(<<p@?mH6v!0#IZ)w3s)uu8Gj+?1=%}z
zB9o$=7v*@4@s-o_(yj{*GAZd6xTG1mxp%NE6PU6f=VdB?<I2fD<y@7wx%-S9FOrAO
zIr*zRZtFvXB{9qT^ciYs&x)Q0D^^rqlFeMPD(0HB=Ty4JV@z<)>vCR^V4dUf290#a
z_r%=LNovU6!8<akURk=SOW_bKt?AO*bm_^&dEA~y0x@H<k9n1hkNfFSW+Md`ob>z2
z0Kt-m{cP8}@3rHW(OKR!(7sM-ir9=1W;8c7;IjC)k+O%OX?24O)@WV@gHa#n1@wgD
zAUQmaS|$sIQHlgcv>GNHG+Y!f-eS7Dh%=^Gs-qW^RuC}E7Zavfwcq#_X$EO#Qd7e(
zvqNKFrO154&^?XYu$2qDQGC=jKBxM$UDNxDO+Gh61mVF^D@ohI2~ol<L16Q$7am5v
zTuNxjr}g5@s8^C<G|*?lqIhVjVK26Jbb3i>)Q)h7dF`tY8YZ-skMCiJxy!BLxx*+m
zE0BN;tb6IYXqdFRKCePCF}DP`(H^z1fQnC37|(OJ!w46h8y2S4oHQK3i;RwZZEKX7
zq#vaUxw=6X^*U<IE0JOdtT_<n@nIu@FclDyba5{pL_h}IPNoNA4A7G%c->aLUA<}3
zzy#4Ddk@sZ-KRbd`qV^&EJ@+O(|6C3Uh-6j+yp4cCdipP!!y7IGB1b+u?d1T(AhQy
zsjN5`pTe@U?A8j)?wm90EI7BERksGohP&+4oVvT_@V7;?_~orA{GoQrI3fb3N1BwH
z<j1b80rj7`l77H&JO;`j<1IQsK^$;opa6ATmg7pU<|Id9j;K~7WV!MK7jsVvfRqAK
zmzS7*8$<3f4aZyAv^pfIg?vm9O_BWu&+dPR!&@9Y3>pEr>~_-EOv}2XPNx_7>zqLU
z=<VR@wEFpUzFvS1-}n~%x3SNCd^4t#6l%_zvj}hBa9WHB#`8s4`4hYhr@{j)Bi1Jq
zS4F5SmM93(Tv<4nlK}0|kwM`hoU7B}w^e}rCIFoR)C~#>5CG62fcLS9YV3lJo9~Q0
zul^sQBG~(_$|}$qPXY8|#-t|Wf5<stxJG*c=+AILzjr`GxG||y3JMFZZ%k&UT#+2=
zgw>VE0sbvlCO0qVSUM}$X)$WzCJM~z1T_d-%w3U|Aw~7UR&`m|IZK>?Y3nFgs}qzU
zEW!+kLC*_Lf{KFra1UR`7*0U@Dxj*A(-Bp9R+99k{PN_b;Qukbq>GLihzzUj?-*<#
zP6Lo_NM2IBT%!^{GNW{GQD+o8cfFl=_q;(nOw;fzFER?Js65-y5JSDsi8yzzAo7;?
z{(;vRw*ZP!y5*fsiPJ5wyJcGy7MI)khPJj)vWo^OF=5ZMA`c?qDEgy;HcudM8k;rf
z9r*XJ4`W7V%yf^Lc`*ajf+1zC=i8b}ai6WCqD<?#c?dX46(b*h^7Bvp;K4`!w*S!s
zi?!Ww82Jo!e-(pG#E50iXEF3|VrZzExt6P0cXYL_jRc^sV&kcTtQNB%4NauQ%T_eY
z6X~r-{qJ!<R=7)tC?>NdRDxY$E(yoE1OzLi_J3NjS(kIyD{|!CSL+Uwad3J;b>9GD
zXj!=|4QzjZ1_Km<q`Kl<bLN~mcgb0D-*Rs{WqhtVb!XmwfZ1y<Xm<m$IxxLz@tYUN
zTUGr2kmvcK;`gowy(0kc7ib2NhR6W`ismz61}s`dWDX{<%OfN^$OLY71(LlfBu|T{
zO^JQ&>n#vpKrE`Q{|hYluXA`GLxbii;_4N|Ch98~Cjet;{?9__qm|-eC1d&uAb0^E
zEo&4o#MLcdf<O+y2^?J=(Ki2@Yv~8|$=<<FwPywo&;=TcJT0(o2Cf1f;W*YkTl6d}
z+uqp$_ELlHoo;+v9wTS~%UA$^lBUMt4QK(}%+YThl6<DVZCuVoJ+M3A9m~$adjNLO
z+aqihq(jVN_GW+K+3>!kebfKMCP2`8m*C`me%ZtVX0&JA@-*?byq$yBu<Ht{y}Ix<
z6nh1sG7C|6zy6{25jZ8<EOR!D;{SlM;6ylRX#e_MP_~oRaM~x!RzX{?!B7)6%oJ^v
z*6wp>TX#k7?`+#%fpM0@8d~wpe?XqC?X*n?a4MYmKca><DaQB@LA`+>)g9H}2sO!W
zV2}!vyNrfmohk1u;ddF|bw^Fk`@2EdkAgr!E(rSQdV6FESwYZFT0!8m?b9~IW*xJ^
z{|v*W)k=Ny7KebVZ&z$_%B8Yfd#6^f-L6#|ixA|dbw;R#FwKWZeBPf=4AE;!GR=Fu
zym$Q_YQ9emeGIP;&wQ@2?<Bl(`z{p^j%?S@lEOat&b0N-e|@g^Y1^ARsCxXK_3@8+
Vf{d89h<<L-okdS)KjaqOuK=r;_fP-;

literal 0
HcmV?d00001

diff --git a/yolo/net/net.py b/yolo/net/net.py
index 005d0de..ac6b7ef 100644
--- a/yolo/net/net.py
+++ b/yolo/net/net.py
@@ -1,3 +1,5 @@
+#!/usr/bin/python 
+# -*- coding: utf-8 -*- 
 from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
@@ -51,7 +53,6 @@ def _variable_with_weight_decay(self, name, shape, stddev, wd, pretrain=True, tr
       stddev: standard devision of a truncated Gaussian
       wd: add L2Loss weight decay multiplied by this float. If None, weight 
       decay is not added for this Variable.
-
    Returns:
       Variable Tensor 
     """
@@ -74,16 +75,19 @@ def conv2d(self, scope, input, kernel_size, stride=1, pretrain=True, train=True)
       output: 4-D tensor [batch_size, height/stride, width/stride, out_channels]
     """
     with tf.variable_scope(scope) as scope:
+      # 初始化权重的kernel
       kernel = self._variable_with_weight_decay('weights', 
                                       shape=kernel_size,
                                       stddev=5e-2,
                                       wd=self.weight_decay, pretrain=pretrain, train=train)
       conv = tf.nn.conv2d(input, kernel, [1, stride, stride, 1], padding='SAME')
+
+      # biases 初始化采用常数 0.0 初始化
       biases = self._variable_on_cpu('biases', kernel_size[3:], tf.constant_initializer(0.0), pretrain, train)
-      bias = tf.nn.bias_add(conv, biases)
-      conv1 = self.leaky_relu(bias)
+      conv1 = tf.nn.bias_add(conv, biases)
+      output = self.leaky_relu(conv1)
 
-    return conv1
+    return output
 
 
   def max_pool(self, input, kernel_size, stride):
@@ -137,6 +141,9 @@ def leaky_relu(self, x, alpha=0.1, dtype=tf.float32):
       y : Tensor
     """
     x = tf.cast(x, dtype=dtype)
+    # 对输入的特征向量进行leaky_relu
+    # 其中对>0的数据采用直接激活的方式，对小于0的数据采用leaky激活方式
+    # 此处实现值得学习和借鉴
     bool_mask = (x > 0)
     mask = tf.cast(bool_mask, dtype=dtype)
     return 1.0 * mask * x + alpha * (1 - mask) * x
@@ -160,4 +167,13 @@ def loss(self, predicts, labels, objects_num):
       labels  : 3-D tensor of [batch_size, max_objects, 5]
       objects_num: 1-D tensor [batch_size]
     """
-    raise NotImplementedError
\ No newline at end of file
+    raise NotImplementedError
+
+'''
+## weight decay：
+在机器学习或者模式识别中，会出现overfitting，而当网络逐渐overfitting时网络
+权值逐渐变大，因此，为了避免出现overfitting,会给误差函数添加一个惩罚项，常用
+的惩罚项是所有权重的平方乘以一个衰减常量之和。其用来惩罚大的权值。
+权值衰减惩罚项使得权值收敛到较小的绝对值，而惩罚大的权值。因为大的权值会使得
+系统出现过拟合，降低其泛化性能。
+'''
\ No newline at end of file
diff --git a/yolo/net/yolo_net.py b/yolo/net/yolo_net.py
index 0dfa034..71abb2c 100644
--- a/yolo/net/yolo_net.py
+++ b/yolo/net/yolo_net.py
@@ -1,3 +1,5 @@
+#!/usr/bin/python 
+# -*- coding: utf-8 -*- 
 from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
@@ -123,6 +125,7 @@ def iou(self, boxes1, boxes2):
     Return:
       iou: 3-D tensor [CELL_SIZE, CELL_SIZE, BOXES_PER_CELL]
     """
+    # 计算左上角和右下角的位置信息
     boxes1 = tf.pack([boxes1[:, :, :, 0] - boxes1[:, :, :, 2] / 2, boxes1[:, :, :, 1] - boxes1[:, :, :, 3] / 2,
                       boxes1[:, :, :, 0] + boxes1[:, :, :, 2] / 2, boxes1[:, :, :, 1] + boxes1[:, :, :, 3] / 2])
     boxes1 = tf.transpose(boxes1, [1, 2, 3, 0])
@@ -134,6 +137,16 @@ def iou(self, boxes1, boxes2):
     rd = tf.minimum(boxes1[:, :, :, 2:], boxes2[2:])
 
     #intersection
+    '''
+    0, 0------------------------>
+    |    ————————————|
+    |    |     ——————|——————    
+    |    |     |     |      |
+    |    |—————|—————|      |
+    |          |____________| 
+    |
+    v
+    '''
     intersection = rd - lu 
 
     inter_square = intersection[:, :, :, 0] * intersection[:, :, :, 1]
@@ -285,7 +298,17 @@ def loss(self, predicts, labels, objects_num):
       label = labels[i, :, :]
       object_num = objects_num[i]
       nilboy = tf.ones([7,7,2])
-      tuple_results = tf.while_loop(self.cond1, self.body1, [tf.constant(0), object_num, [class_loss, object_loss, noobject_loss, coord_loss], predict, label, nilboy])
+      tuple_results = tf.while_loop(
+                        self.cond1, 
+                        self.body1, 
+                        [
+                            tf.constant(0), 
+                            object_num, 
+                            [class_loss, object_loss, noobject_loss, coord_loss], 
+                            predict, 
+                            label, 
+                            nilboy
+                        ])
       for j in range(4):
         loss[j] = loss[j] + tuple_results[2][j]
       nilboy = tuple_results[5]
diff --git a/yolo/net/yolo_tiny_net.py b/yolo/net/yolo_tiny_net.py
index 6f3c48e..11d24ea 100644
--- a/yolo/net/yolo_tiny_net.py
+++ b/yolo/net/yolo_tiny_net.py
@@ -1,3 +1,5 @@
+#!/usr/bin/python 
+# -*- coding: utf-8 -*- 
 from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
@@ -117,7 +119,10 @@ def iou(self, boxes1, boxes2):
                       boxes2[0] + boxes2[2] / 2, boxes2[1] + boxes2[3] / 2])
 
     #calculate the left up point
+    # boxes相当于grandtruth的box,只有一个
+    # 但是boxes1相当于ROI，是有很多的
     lu = tf.maximum(boxes1[:, :, :, 0:2], boxes2[0:2])
+    # calculate the right down point
     rd = tf.minimum(boxes1[:, :, :, 2:], boxes2[2:])
 
     #intersection
@@ -135,13 +140,14 @@ def iou(self, boxes1, boxes2):
     
     return inter_square/(square1 + square2 - inter_square + 1e-6)
 
+  # loop停止函数和后面的运行函数体的输入参数是一致的，输入相同的参数数据
   def cond1(self, num, object_num, loss, predict, label, nilboy):
     """
     if num < object_num
     """
     return num < object_num
 
-
+  # 运行的函数体的定义
   def body1(self, num, object_num, loss, predict, labels, nilboy):
     """
     calculate loss
@@ -202,10 +208,10 @@ def body1(self, num, object_num, loss, predict, labels, nilboy):
 
     p_C = predict[:, :, self.num_classes:self.num_classes + self.boxes_per_cell]
 
-    #calculate truth x,y,sqrt_w,sqrt_h 0-D
+    #calculate truth x, y, sqrt_w, sqrt_h 0-D
     x = label[0]
     y = label[1]
-
+    # TODO:为啥要使用sqrt的宽度和高度来
     sqrt_w = tf.sqrt(tf.abs(label[2]))
     sqrt_h = tf.sqrt(tf.abs(label[3]))
     #sqrt_w = tf.abs(label[2])
@@ -223,6 +229,7 @@ def body1(self, num, object_num, loss, predict, labels, nilboy):
     #p_sqrt_h = predict_boxes[:, :, :, 3]
     p_sqrt_w = tf.sqrt(tf.minimum(self.image_size * 1.0, tf.maximum(0.0, predict_boxes[:, :, :, 2])))
     p_sqrt_h = tf.sqrt(tf.minimum(self.image_size * 1.0, tf.maximum(0.0, predict_boxes[:, :, :, 3])))
+    
     #calculate truth p 1-D tensor [NUM_CLASSES]
     P = tf.one_hot(tf.cast(label[4], tf.int32), self.num_classes, dtype=tf.float32)
 
@@ -262,17 +269,27 @@ def loss(self, predicts, labels, objects_num):
       labels  : 3-D tensor of [batch_size, max_objects, 5]
       objects_num: 1-D tensor [batch_size]
     """
-    class_loss = tf.constant(0, tf.float32)
-    object_loss = tf.constant(0, tf.float32)
-    noobject_loss = tf.constant(0, tf.float32)
-    coord_loss = tf.constant(0, tf.float32)
+    # 定义不同的loss变量
+    class_loss = tf.constant(0, tf.float32)       # 分类损失
+    object_loss = tf.constant(0, tf.float32)      # 有对象的时候与ground truth的损失
+    noobject_loss = tf.constant(0, tf.float32)    # 预测框中没有对象的时候的损失
+    coord_loss = tf.constant(0, tf.float32)       # 
+
     loss = [0, 0, 0, 0]
     for i in range(self.batch_size):
       predict = predicts[i, :, :, :]
       label = labels[i, :, :]
       object_num = objects_num[i]
       nilboy = tf.ones([7,7,2])
-      tuple_results = tf.while_loop(self.cond1, self.body1, [tf.constant(0), object_num, [class_loss, object_loss, noobject_loss, coord_loss], predict, label, nilboy])
+      tuple_results = tf.while_loop(self.cond1,# 其输入参数就是后面list中的变量
+                                    self.body1,# 输入参数就是其后的list变量，详细见其定义 
+                                    [  tf.constant(0), 
+                                       object_num, 
+                                       [class_loss, object_loss, noobject_loss, coord_loss], 
+                                       predict, 
+                                       label, 
+                                       nilboy
+                                    ])
       for j in range(4):
         loss[j] = loss[j] + tuple_results[2][j]
       nilboy = tuple_results[5]
diff --git a/yolo/solver/solver.py b/yolo/solver/solver.py
index 50bb8d4..8fd65b3 100644
--- a/yolo/solver/solver.py
+++ b/yolo/solver/solver.py
@@ -1,3 +1,6 @@
+#!/usr/bin/python 
+# -*- coding: utf-8 -*- 
+
 """Solver Abstract class 
 """
 class Solver(object):
diff --git a/yolo/solver/yolo_solver.py b/yolo/solver/yolo_solver.py
index 0d9b4f3..6888480 100644
--- a/yolo/solver/yolo_solver.py
+++ b/yolo/solver/yolo_solver.py
@@ -1,3 +1,5 @@
+#!/usr/bin/python 
+# -*- coding: utf-8 -*- 
 from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
@@ -54,6 +56,7 @@ def _train(self):
   def construct_graph(self):
     # construct graph
     self.global_step = tf.Variable(0, trainable=False)
+    # 搭建神经网络模型的输入和输出结构
     self.images = tf.placeholder(tf.float32, (self.batch_size, self.height, self.width, 3))
     self.labels = tf.placeholder(tf.float32, (self.batch_size, self.max_objects, 5))
     self.objects_num = tf.placeholder(tf.int32, (self.batch_size))
@@ -71,6 +74,7 @@ def solve(self):
 
     init =  tf.global_variables_initializer()
 
+    # Merges all summaries collected in the default graph.
     summary_op = tf.summary.merge_all()
 
     sess = tf.Session()
@@ -83,12 +87,15 @@ def solve(self):
 
     for step in xrange(self.max_iterators):
       start_time = time.time()
+      # 获取train data
       np_images, np_labels, np_objects_num = self.dataset.batch()
-
-      _, loss_value, nilboy = sess.run([self.train_op, self.total_loss, self.nilboy], feed_dict={self.images: np_images, self.labels: np_labels, self.objects_num: np_objects_num})
+      # 训练模型一个batch
+      _, loss_value, nilboy = sess.run([self.train_op, self.total_loss, self.nilboy], 
+                                      feed_dict= { self.images: np_images, 
+                                                   self.labels: np_labels, 
+                                                   self.objects_num: np_objects_num
+                                                 })
       #loss_value, nilboy = sess.run([self.total_loss, self.nilboy], feed_dict={self.images: np_images, self.labels: np_labels, self.objects_num: np_objects_num})
-
-
       duration = time.time() - start_time
 
       assert not np.isnan(loss_value), 'Model diverged with loss = NaN'
@@ -105,7 +112,11 @@ def solve(self):
 
         sys.stdout.flush()
       if step % 100 == 0:
-        summary_str = sess.run(summary_op, feed_dict={self.images: np_images, self.labels: np_labels, self.objects_num: np_objects_num})
+        summary_str = sess.run(summary_op, feed_dict={
+                                                self.images: np_images, 
+                                                self.labels: np_labels, 
+                                                self.objects_num: np_objects_num
+                                                })
         summary_writer.add_summary(summary_str, step)
       if step % 5000 == 0:
         saver2.save(sess, self.train_dir + '/model.ckpt', global_step=step)
diff --git a/yolo/utils/process_config.py b/yolo/utils/process_config.py
index 93a783f..a8fbbb0 100644
--- a/yolo/utils/process_config.py
+++ b/yolo/utils/process_config.py
@@ -1,3 +1,9 @@
+#!/usr/bin/python 
+# -*- coding: utf-8 -*- 
+"""
+从配置文件中读取相应的配置参数设置
+这个方法是可以重用的
+"""
 import ConfigParser
 
 def process_config(conf_file):

From 6e543b128d77f9f4f6391fe242fc4bd612d6460b Mon Sep 17 00:00:00 2001
From: wenlong <wenlong@nfs.iscas.ac.cn>
Date: Mon, 13 Mar 2017 16:23:46 +0800
Subject: [PATCH 2/4] update a .gitignore

---
 .gitignore        |   2 ++
 yolo/__init__.pyc | Bin 203 -> 0 bytes
 2 files changed, 2 insertions(+)
 delete mode 100644 yolo/__init__.pyc

diff --git a/.gitignore b/.gitignore
index 0d20b64..1287439 100644
--- a/.gitignore
+++ b/.gitignore
@@ -1 +1,3 @@
 *.pyc
+*/*.pyc
+*/*/*/.pyc
diff --git a/yolo/__init__.pyc b/yolo/__init__.pyc
deleted file mode 100644
index 6eebee04818f9289749078a326b82d3013973776..0000000000000000000000000000000000000000
GIT binary patch
literal 0
HcmV?d00001

literal 203
zcmZSn%*!<^;%G!N0~9a;X$K%K)&LSIK*Y$9!@v*)XEQQHF~Y=|7^0X!3W7C2x-$O*
zfu9C5kXgbG(wtb5Se#l?0usmr(riGwI6tQ>wFqQ11A+is<!lublv<D<Q<9ojoL`id
vlV7e|nV*v%1193*GxIV_;^XxSDoZ$kdTny^Q%ZAE?HEBuvH%GlMjl20x63Q_


From d7f91c5ac34905768e55801138c9529aefe59327 Mon Sep 17 00:00:00 2001
From: wenlong <wenlong@nfs.iscas.ac.cn>
Date: Fri, 17 Mar 2017 09:01:47 +0800
Subject: [PATCH 3/4] <description> 1. add some reading comment for review
 </description>

---
 yolo/net/net.py            |  4 ++--
 yolo/net/yolo_tiny_net.py  | 17 ++++++++++++++---
 yolo/solver/yolo_solver.py | 14 +++++++++-----
 3 files changed, 25 insertions(+), 10 deletions(-)

diff --git a/yolo/net/net.py b/yolo/net/net.py
index ac6b7ef..d731004 100644
--- a/yolo/net/net.py
+++ b/yolo/net/net.py
@@ -103,7 +103,7 @@ def max_pool(self, input, kernel_size, stride):
     return tf.nn.max_pool(input, ksize=[1, kernel_size[0], kernel_size[1], 1], strides=[1, stride, stride, 1],
                   padding='SAME')
 
-  def local(self, scope, input, in_dimension, out_dimension, leaky=True, pretrain=True, train=True):
+  def local(self, scope, _input, in_dimension, out_dimension, leaky=True, pretrain=True, train=True):
     """Fully connection layer
 
     Args:
@@ -114,7 +114,7 @@ def local(self, scope, input, in_dimension, out_dimension, leaky=True, pretrain=
       output: 2-D tensor [batch_size, out_dimension]
     """
     with tf.variable_scope(scope) as scope:
-      reshape = tf.reshape(input, [tf.shape(input)[0], -1])
+      reshape = tf.reshape(_input, [tf.shape(_input)[0], -1])
 
       weights = self._variable_with_weight_decay('weights', shape=[in_dimension, out_dimension],
                           stddev=0.04, wd=self.weight_decay, pretrain=pretrain, train=train)
diff --git a/yolo/net/yolo_tiny_net.py b/yolo/net/yolo_tiny_net.py
index 11d24ea..696790a 100644
--- a/yolo/net/yolo_tiny_net.py
+++ b/yolo/net/yolo_tiny_net.py
@@ -41,7 +41,16 @@ def inference(self, images):
       predicts: 4-D tensor [batch_size, cell_size, cell_size, num_classes + 5 * boxes_per_cell]
     """
     conv_num = 1
-
+    """
+    conv2d(self, scope, input, kernel_size, stride=1, pretrain=True, train=True)
+    Args:
+      input: 4-D tensor [batch_size, height, width, depth]
+      scope: variable_scope name 
+      kernel_size: [k_height, k_width, in_channel, out_channel]
+      stride: int32
+    Return:
+      output: 4-D tensor [batch_size, height/stride, width/stride, out_channels]
+    """
     temp_conv = self.conv2d('conv' + str(conv_num), images, [3, 3, 3, 16], stride=1)
     conv_num += 1
 
@@ -98,7 +107,7 @@ def inference(self, images):
     scales = tf.reshape(local3[:, n1:n2], (-1, self.cell_size, self.cell_size, self.boxes_per_cell))
     boxes = tf.reshape(local3[:, n2:], (-1, self.cell_size, self.cell_size, self.boxes_per_cell * 4))
 
-    local3 = tf.concat([class_probs, scales, boxes], 3)
+    local3 = tf.concat(3, [class_probs, scales, boxes])
 
     predicts = local3
 
@@ -273,7 +282,7 @@ def loss(self, predicts, labels, objects_num):
     class_loss = tf.constant(0, tf.float32)       # 分类损失
     object_loss = tf.constant(0, tf.float32)      # 有对象的时候与ground truth的损失
     noobject_loss = tf.constant(0, tf.float32)    # 预测框中没有对象的时候的损失
-    coord_loss = tf.constant(0, tf.float32)       # 
+    coord_loss = tf.constant(0, tf.float32)       # 预测框位置信息损失
 
     loss = [0, 0, 0, 0]
     for i in range(self.batch_size):
@@ -281,6 +290,7 @@ def loss(self, predicts, labels, objects_num):
       label = labels[i, :, :]
       object_num = objects_num[i]
       nilboy = tf.ones([7,7,2])
+      # 返回值就是被调用函数题的输入数据
       tuple_results = tf.while_loop(self.cond1,# 其输入参数就是后面list中的变量
                                     self.body1,# 输入参数就是其后的list变量，详细见其定义 
                                     [  tf.constant(0), 
@@ -290,6 +300,7 @@ def loss(self, predicts, labels, objects_num):
                                        label, 
                                        nilboy
                                     ])
+      #累加各类loss值
       for j in range(4):
         loss[j] = loss[j] + tuple_results[2][j]
       nilboy = tuple_results[5]
diff --git a/yolo/solver/yolo_solver.py b/yolo/solver/yolo_solver.py
index 6888480..9a747e7 100644
--- a/yolo/solver/yolo_solver.py
+++ b/yolo/solver/yolo_solver.py
@@ -91,15 +91,17 @@ def solve(self):
       np_images, np_labels, np_objects_num = self.dataset.batch()
       # 训练模型一个batch
       _, loss_value, nilboy = sess.run([self.train_op, self.total_loss, self.nilboy], 
-                                      feed_dict= { self.images: np_images, 
-                                                   self.labels: np_labels, 
-                                                   self.objects_num: np_objects_num
-                                                 })
+                                        feed_dict= { 
+                                                  self.images: np_images, 
+                                                  self.labels: np_labels, 
+                                                  self.objects_num: np_objects_num 
+                                                  })
       #loss_value, nilboy = sess.run([self.total_loss, self.nilboy], feed_dict={self.images: np_images, self.labels: np_labels, self.objects_num: np_objects_num})
-      duration = time.time() - start_time
 
+      duration = time.time() - start_time
       assert not np.isnan(loss_value), 'Model diverged with loss = NaN'
 
+      #10次迭代输入计算信息
       if step % 10 == 0:
         num_examples_per_step = self.dataset.batch_size
         examples_per_sec = num_examples_per_step / duration
@@ -111,6 +113,7 @@ def solve(self):
                              examples_per_sec, sec_per_batch))
 
         sys.stdout.flush()
+      # 100次迭代更新后将预测结果写入文件
       if step % 100 == 0:
         summary_str = sess.run(summary_op, feed_dict={
                                                 self.images: np_images, 
@@ -118,6 +121,7 @@ def solve(self):
                                                 self.objects_num: np_objects_num
                                                 })
         summary_writer.add_summary(summary_str, step)
+      # 5000次迭代保存一个模型
       if step % 5000 == 0:
         saver2.save(sess, self.train_dir + '/model.ckpt', global_step=step)
     sess.close()

From 069f567488051782dea0eb6bf093d719abba471d Mon Sep 17 00:00:00 2001
From: cwlseu <caowenlong92@gmail.com>
Date: Fri, 24 Mar 2017 11:45:31 +0800
Subject: [PATCH 4/4] =?UTF-8?q?<dev>=20=E9=80=9A=E8=BF=87=E9=98=85?=
 =?UTF-8?q?=E8=AF=BByolo=E4=BB=A3=E7=A0=81=EF=BC=8C=E6=B7=BB=E5=8A=A0?=
 =?UTF-8?q?=E4=BA=86=E6=B3=A8=E9=87=8A=20</dev>?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 yolo/dataset/__init__.pyc     | Bin 195 -> 0 bytes
 yolo/dataset/dataset.pyc      | Bin 886 -> 0 bytes
 yolo/dataset/text_dataset.pyc | Bin 4740 -> 0 bytes
 3 files changed, 0 insertions(+), 0 deletions(-)
 delete mode 100644 yolo/dataset/__init__.pyc
 delete mode 100644 yolo/dataset/dataset.pyc
 delete mode 100644 yolo/dataset/text_dataset.pyc

diff --git a/yolo/dataset/__init__.pyc b/yolo/dataset/__init__.pyc
deleted file mode 100644
index 7dec5bd23b663c719b65ec04b66790e46ea832bb..0000000000000000000000000000000000000000
GIT binary patch
literal 0
HcmV?d00001

literal 195
zcmZSn%*!<^;%G!N0~9a;X$K%KmH`qeK*Y$9!@v*)XEQQHF@gkxH9#se{{w*^m|4OO
z(vVn^Se#l?!UJTNq*j!~Lj^Q|d<FzjYyjjsTg3#W7Uaj2q~;ap7p3Lom+MyM=j6wL
n38=o9`1s7c%#!$cy@JXT4xlER-29Z%oK!nTAiEgk0v<*H(z7l{

diff --git a/yolo/dataset/dataset.pyc b/yolo/dataset/dataset.pyc
deleted file mode 100644
index 14f4f69b06f043d16c5c102cd74142947f1ec49d..0000000000000000000000000000000000000000
GIT binary patch
literal 0
HcmV?d00001

literal 886
zcmb_a%TB{E5ZokfN+T5xh`W!u0QU+ZXhEo#9@0yRB3WtTilQWr;zg)e;4AnazK9=S
z*A0(KZ~-ED;%vq<v$nn0LGV0(nP<>G9=*?K7(tK&#Doqg2bK=xE<6D|0$j%~<Q~kz
z4l@r3E>cWq3WZptMv1IQjS&I%KL30eI8+<)GC`oofOH8-&|&r1Yi;5VM6;0Nm_33K
z>$Wa|5E=jA!E$37CLln72a<Crnai~a0g;KW%TiaePMfqeLvb!PGm)<{6qb=w14iNZ
z6w8=~?1lRTQ;n0dE>x*1RQYAo=*A{RyC14p;(%CNrnpO$ZIh5HV8I;IY&=XNRqF&*
zWpuMF^!?#l7do-fu1wnZQN6|;S|jDETA`Fa%?-B<-HcPdj~{o2(_GPcEmF*GzO)zd
zO513}a(|52B?;@m+p)>#g<NdIS81uFL_RtxOP${p{K>aQ%3NnsMof;lyU!%Q-m!>e
p?42<z=)1nt?@jHwMSUtDSM}mnWoViI;Aq=C*)(=)ZzO~7yaC*+#QgvO

diff --git a/yolo/dataset/text_dataset.pyc b/yolo/dataset/text_dataset.pyc
deleted file mode 100644
index 7f55d47836bcf79401a8d925707657bed9d1e989..0000000000000000000000000000000000000000
GIT binary patch
literal 0
HcmV?d00001

literal 4740
zcmb_f-EtdQ5$+lN%l|)d>=-svOBGRSy(wbv0_6=XNlfil0bZDFk*qdmb}WsoaU_jA
za}>*zbrtTZ;*u)fAy>c)aKq(30T(;~RXhOueSJo@6PBALmUYxW=bY)&r~B)k!oTP0
ze|++{CoReU3i$oW7mj3q2GPd9BR!G6BTYvf@~-l(G+oOVlrKoLVELl-iqb6VTF@nF
zmZo%Bn&l~7k!EE|SEX4~x+J~2^y|{B+r7%to00ykG-oYek=~s2=cPFhzADkYbR22d
z<aZ)3L_TXSNV_PH8g-Z+^E?d)gIT~39%e~zWTGJM50lhbMHPy6{5;O$WMC!pkPOp!
zV1mwQ&@zzkTSW<qUyY(sWZ5#<PrGRpwk=-)zZaVDQDn>vb<x=5{Do&}%L97$0}zqW
zq3|c%2p8i3@E?3j0=_FK0ml`UfY(Y&RK$PO;KQ>&y!w>7=MB@O6=j)c@N6$03`fT6
z#J$LC!>%kc9;T>Dg`Fhrhh`Tt&x`xvQ4|bA)Ah#vcyQM{({bDn$D}#7dSRB`RrR*2
zw?T(v(i~O2jm=sNAG)9ZddS1R#IcRYxgK-lg(K%8fg{G1bDZCmsi^SZqT*CkxR_N^
z5%wxxh|1Ghr3+C-OjX)&S_NU@@PW0dq&k(<<p@?mH6v!0#IZ)w3s)uu8Gj+?1=%}z
zB9o$=7v*@4@s-o_(yj{*GAZd6xTG1mxp%NE6PU6f=VdB?<I2fD<y@7wx%-S9FOrAO
zIr*zRZtFvXB{9qT^ciYs&x)Q0D^^rqlFeMPD(0HB=Ty4JV@z<)>vCR^V4dUf290#a
z_r%=LNovU6!8<akURk=SOW_bKt?AO*bm_^&dEA~y0x@H<k9n1hkNfFSW+Md`ob>z2
z0Kt-m{cP8}@3rHW(OKR!(7sM-ir9=1W;8c7;IjC)k+O%OX?24O)@WV@gHa#n1@wgD
zAUQmaS|$sIQHlgcv>GNHG+Y!f-eS7Dh%=^Gs-qW^RuC}E7Zavfwcq#_X$EO#Qd7e(
zvqNKFrO154&^?XYu$2qDQGC=jKBxM$UDNxDO+Gh61mVF^D@ohI2~ol<L16Q$7am5v
zTuNxjr}g5@s8^C<G|*?lqIhVjVK26Jbb3i>)Q)h7dF`tY8YZ-skMCiJxy!BLxx*+m
zE0BN;tb6IYXqdFRKCePCF}DP`(H^z1fQnC37|(OJ!w46h8y2S4oHQK3i;RwZZEKX7
zq#vaUxw=6X^*U<IE0JOdtT_<n@nIu@FclDyba5{pL_h}IPNoNA4A7G%c->aLUA<}3
zzy#4Ddk@sZ-KRbd`qV^&EJ@+O(|6C3Uh-6j+yp4cCdipP!!y7IGB1b+u?d1T(AhQy
zsjN5`pTe@U?A8j)?wm90EI7BERksGohP&+4oVvT_@V7;?_~orA{GoQrI3fb3N1BwH
z<j1b80rj7`l77H&JO;`j<1IQsK^$;opa6ATmg7pU<|Id9j;K~7WV!MK7jsVvfRqAK
zmzS7*8$<3f4aZyAv^pfIg?vm9O_BWu&+dPR!&@9Y3>pEr>~_-EOv}2XPNx_7>zqLU
z=<VR@wEFpUzFvS1-}n~%x3SNCd^4t#6l%_zvj}hBa9WHB#`8s4`4hYhr@{j)Bi1Jq
zS4F5SmM93(Tv<4nlK}0|kwM`hoU7B}w^e}rCIFoR)C~#>5CG62fcLS9YV3lJo9~Q0
zul^sQBG~(_$|}$qPXY8|#-t|Wf5<stxJG*c=+AILzjr`GxG||y3JMFZZ%k&UT#+2=
zgw>VE0sbvlCO0qVSUM}$X)$WzCJM~z1T_d-%w3U|Aw~7UR&`m|IZK>?Y3nFgs}qzU
zEW!+kLC*_Lf{KFra1UR`7*0U@Dxj*A(-Bp9R+99k{PN_b;Qukbq>GLihzzUj?-*<#
zP6Lo_NM2IBT%!^{GNW{GQD+o8cfFl=_q;(nOw;fzFER?Js65-y5JSDsi8yzzAo7;?
z{(;vRw*ZP!y5*fsiPJ5wyJcGy7MI)khPJj)vWo^OF=5ZMA`c?qDEgy;HcudM8k;rf
z9r*XJ4`W7V%yf^Lc`*ajf+1zC=i8b}ai6WCqD<?#c?dX46(b*h^7Bvp;K4`!w*S!s
zi?!Ww82Jo!e-(pG#E50iXEF3|VrZzExt6P0cXYL_jRc^sV&kcTtQNB%4NauQ%T_eY
z6X~r-{qJ!<R=7)tC?>NdRDxY$E(yoE1OzLi_J3NjS(kIyD{|!CSL+Uwad3J;b>9GD
zXj!=|4QzjZ1_Km<q`Kl<bLN~mcgb0D-*Rs{WqhtVb!XmwfZ1y<Xm<m$IxxLz@tYUN
zTUGr2kmvcK;`gowy(0kc7ib2NhR6W`ismz61}s`dWDX{<%OfN^$OLY71(LlfBu|T{
zO^JQ&>n#vpKrE`Q{|hYluXA`GLxbii;_4N|Ch98~Cjet;{?9__qm|-eC1d&uAb0^E
zEo&4o#MLcdf<O+y2^?J=(Ki2@Yv~8|$=<<FwPywo&;=TcJT0(o2Cf1f;W*YkTl6d}
z+uqp$_ELlHoo;+v9wTS~%UA$^lBUMt4QK(}%+YThl6<DVZCuVoJ+M3A9m~$adjNLO
z+aqihq(jVN_GW+K+3>!kebfKMCP2`8m*C`me%ZtVX0&JA@-*?byq$yBu<Ht{y}Ix<
z6nh1sG7C|6zy6{25jZ8<EOR!D;{SlM;6ylRX#e_MP_~oRaM~x!RzX{?!B7)6%oJ^v
z*6wp>TX#k7?`+#%fpM0@8d~wpe?XqC?X*n?a4MYmKca><DaQB@LA`+>)g9H}2sO!W
zV2}!vyNrfmohk1u;ddF|bw^Fk`@2EdkAgr!E(rSQdV6FESwYZFT0!8m?b9~IW*xJ^
z{|v*W)k=Ny7KebVZ&z$_%B8Yfd#6^f-L6#|ixA|dbw;R#FwKWZeBPf=4AE;!GR=Fu
zym$Q_YQ9emeGIP;&wQ@2?<Bl(`z{p^j%?S@lEOat&b0N-e|@g^Y1^ARsCxXK_3@8+
Vf{d89h<<L-okdS)KjaqOuK=r;_fP-;