def __getUserCallBack(loginName, password, clientData, httpcode, data, headers, success, url):
"""
参数1：HTTP请求返回码 这个结果集合参数是一个整形值
参数2：返回的内容 它是一个字符串
参数3：服务器返回的HTTP协议头，如：{"Content-Type": "application/x-www-form-urlencoded"} 它是一个字典
参数4：执行是否成功，当请求执行有错误时为False 可以通过httpcode进一步判断错误信息
参数5：请求所用的网址
"""
INFO_MSG('__getUserCallBack: loginName=%s, httpcode=%s, success=%s' % (loginName, httpcode, success))
# 如果获取微信信息失败
if not success:
KBEngine.accountLoginResponse(loginName, loginName, clientData, KBEngine.SERVER_ERR_USER1)
return
# 数据转换
data = GTools.json_load(data)
# 获取用户信息失败
if 'errcode' in data and data['errcode'] != 0:
ERROR_MSG('__getUserCallBack: data=%s' % (data))
KBEngine.accountLoginResponse(loginName, loginName, clientData, KBEngine.SERVER_ERR_USER1)
return
# 检查加密数据
key = GTools.getSha1Str(GTools.byteToStr(clientData) + data['session_key'])
ERROR_MSG("clientData(%s) session_key(%s) key(%s) password(%s)" % (clientData, data['session_key'], key, password))
if key == password:
INFO_MSG("__getUserCallBack::SERVER_SUCCESS. loginName=%s, openid=%s" % (loginName, data['openid']))
KBEngine.accountLoginResponse(loginName, data['openid'], clientData, KBEngine.SERVER_SUCCESS)
else:
ERROR_MSG("__getUserCallBack::SERVER_ERR_NAME_PASSWORD. loginName=%s, openid=%s" % (loginName, data['openid']))
KBEngine.accountLoginResponse(loginName, loginName, clientData, KBEngine.SERVER_ERR_NAME_PASSWORD)
def onClientEnabled(self):
"""
KBEngine method.
该entity被正式激活为可使用， 此时entity已经建立了client对应实体， 可以在此创建它的
cell部分。
"""
DEBUG_MSG("Avatar(%i_%i):onClientEnabled." % (self.id, self.avatarID))
# 更新最后登陆时间
self.lastReloginTime = GTools.nowTime()
self._updateAvatar()
# 获取主大厅失败
mainHalls = GTools.getBaseData_MainHalls()
if mainHalls is None:
ERROR_MSG("Avatar(%i_%i):onClientEnabled. mainHalls is None." %
(self.id, self.avatarID))
return
# 通知主大厅上线
mainHalls.B_tellOnline(self, self.avatarID)
# 当前游戏厅为空 则加入游戏大厅
if self.nowHalls is None:
hallsName = GServerCfg.GC_ServerIDToHalls[self.serverID]
mainHalls.B_reqEnterGameHalls(self, self.avatarID, hallsName)
# 当前游戏厅不为空 则通知当前所在大厅
elif self.client is not None:
self.client.Exs_tellCurHalls(self.nowHallsName)
# 通知在线
if self.cell is not None:
self.cell.C_tellOnlineStatus(True)
def __init__(self):
KBEngine.Entity.__init__(self)
# 已经注册的大厅信息
self.hallsDict = {}
# 在线的角色字典
self.avatarsDict = {}
# 注册自己到base层的全局
GTools.setBaseData_MainHalls(self)
def _onAvatarSaved(self, bSuccess, newAV):
"""
角色实体存库后的回调
参数1：成功或失败
参数2：base实体
"""
# 如果创建过程中账号已经销毁
if self.isDestroyed and newAV:
newAV.destroy(True)
return
if bSuccess:
# 保存实体的唯一ID
newAV.avatarID = newAV.databaseID
# 记录到列表
props = GTools.dcopy(GDataTmp.Avatar_Info_In_Account)
props['avatarID'] = newAV.avatarID
props['serverID'] = newAV.serverID
self.avatarDict[newAV.serverID] = props
self.writeToDB()
# 销毁实体
newAV.destroy()
# 通知创建结果
if self.client is not None:
self.client.Exs_ackCreateAvatar(bSuccess, props)
def __init__(self):
KBEngine.Proxy.__init__(self)
# 当前角色
self.curAvatar = None
# 登陆时客户端的数据
self.clientData = GTools.json_load(self.getClientDatas()[0])
def onClientEnabled(self):
"""
KBEngine method.
该entity被正式激活为可使用， 此时entity已经建立了client对应实体， 可以在此创建它的
cell部分。
"""
INFO_MSG("Account(%i) entities enable." % (self.id))
# 更新最后登陆时间
self.lastReloginTime = GTools.nowTime()
def resetTimer(self, initialOffset):
"""
重置游戏定时器
参数1：时间间隔
"""
# 如果当前有定时器 先关闭
if self.timerID != 0:
self.closeTimer()
# 开始定时器
self.timerID = self.addTimer(initialOffset, 0, 0)
self.startTime = GTools.nowTime()
def onClientDeath(self):
"""
KBEngine method.
客户端对应实体已经销毁
"""
DEBUG_MSG("Avatar(%i_%i):onClientDeath." % (self.id, self.avatarID))
# 通知主大厅上线
mainHalls = GTools.getBaseData_MainHalls()
if mainHalls:
mainHalls.B_tellOffline(self, self.avatarID)
# 通知下线
if self.cell is not None:
self.cell.C_tellOnlineStatus(False)
def onTimer(self, tid, userArg):
"""
KBEngine method.
使用addTimer后， 当时间到达则该接口被调用
@param tid : addTimer 的返回值ID
@param userArg : addTimer 最后一个参数所给入的数据
"""
DEBUG_MSG("%s(%i):onTimer. tid(%s) userArg(%s)." %
(self.hallsName, self.id, tid, userArg))
# 向主大厅上报自己信息 上报完成后停止
if userArg == GDefine.GC_HallsTime['regToMainHalls']['userArg']:
mainHalls = GTools.getBaseData_MainHalls()
if mainHalls:
mainHalls.B_tellHallsInfo(self, self.hallsName)
self.delTimer(tid)
def onRequestAccountLogin(loginName, password, datas):
"""
KBEngine method.
请求登陆账号回调
@param loginName: 客户端请求时所提交的名称 - 微信登陆获得的code
@type loginName: string
@param password: 密码 - sha1( datas + sessionkey )
@type password: string
@param datas: 客户端请求时所附带的数据，可将数据转发第三方平台 - 用户信息字符串
@type datas: bytes
"""
INFO_MSG('onRequestAccountLogin: loginName=%s' % (loginName))
# 此处可通过http等手段将请求提交至第三方平台，平台返回的数据也可放入datas
# datas将会回调至客户端
# 如果使用http访问，因为interfaces是单线程的，同步http访问容易卡住主线程，建议使用
# KBEngine.registerReadFileDescriptor()和KBEngine.registerWriteFileDescriptor()结合
# KBEngine.urlopen("https://www.baidu.com",onHttpCallback)异步访问。也可以结合socket的方式与平台交互。
# 如果返回码为KBEngine.SERVER_ERR_LOCAL_PROCESSING 则表示验证登陆成功，但dbmgr需要检查账号密码
# KBEngine.SERVER_SUCCESS 则无需再检查密码
loginNameList = loginName.split(':')
# 0 登陆方式为账号登陆
if loginNameList[0] == '0':
KBEngine.accountLoginResponse(loginName, loginName, datas, KBEngine.SERVER_ERR_LOCAL_PROCESSING)
# 1 登陆方式为微信登陆
elif loginNameList[0] == '1':
# 换取用户唯一标识 OpenID 和会话密钥 session_key
url = "https://api.weixin.qq.com/sns/jscode2session?appid=%s&secret=%s&js_code=%s&grant_type=authorization_code" \
% (GServerCfg.GC_APPInfo['appID'], GServerCfg.GC_APPInfo['AppSecret'], loginNameList[1])
KBEngine.urlopen(url, GTools.functor(__getUserCallBack, loginName, password, datas))
def setup_inputs(sess, filenames, image_size=None, capacity_factor=3):
if image_size is None:
image_size = FLAGS.sample_size
#pdb.set_trace()
reader = tf.TFRecordReader()
filename_queue = tf.train.string_input_producer(filenames)
key, value = reader.read(filename_queue)
AlsoLabel=True
kKick= myParams.myDict['InputMode'] == 'kKick'
if kKick or myParams.myDict['InputMode'] == '1DFTx' or myParams.myDict['InputMode'] == '1DFTy' or myParams.myDict['InputMode'] == '2DFT':
AlsoLabel=False
batch_size=myParams.myDict['batch_size']
channelsIn=myParams.myDict['channelsIn']
channelsOut=myParams.myDict['channelsOut']
DataH=myParams.myDict['DataH']
DataW=myParams.myDict['DataW']
LabelsH=myParams.myDict['LabelsH']
LabelsW=myParams.myDict['LabelsW']
if myParams.myDict['InputMode'] == 'AAA':
#filename_queue = tf.Print(filename_queue,[filename_queue,],message='ZZZZZZZZZ:')
keyX=key
value = tf.Print(value,[keyX,],message='QQQ:')
featuresA = tf.parse_single_example(
value,
features={
'CurIs': tf.FixedLenFeature([], tf.string),
'Labels': tf.FixedLenFeature([], tf.string)
})
feature = tf.decode_raw(featuresA['Labels'], tf.float32)
CurIs = tf.decode_raw(featuresA['CurIs'], tf.float32)
CurIs = tf.cast(CurIs, tf.int64)
mx=CurIs
feature = tf.Print(feature,[keyX,mx],message='QQQ:')
feature = tf.Print(feature,[keyX,mx],message='QQQ:')
feature = tf.Print(feature,[keyX,mx],message='QQQ:')
feature = tf.Print(feature,[keyX,mx],message='QQQ:')
feature = tf.Print(feature,[keyX,mx],message='QQQ:')
feature = tf.reshape(feature, [DataH, DataW, channelsIn])
feature = tf.cast(feature, tf.float32)
label=feature
features, labels = tf.train.batch([feature, label],
batch_size=batch_size,
num_threads=4,
capacity = capacity_factor*batch_size,
name='labels_and_features')
tf.train.start_queue_runners(sess=sess)
return features, labels
#image = tf.image.decode_jpeg(value, channels=channels, name="dataset_image")
#print('1')
if AlsoLabel:
featuresA = tf.parse_single_example(
value,
features={
'DataH': tf.FixedLenFeature([], tf.int64),
'DataW': tf.FixedLenFeature([], tf.int64),
'channelsIn': tf.FixedLenFeature([], tf.int64),
'LabelsH': tf.FixedLenFeature([], tf.int64),
'LabelsW': tf.FixedLenFeature([], tf.int64),
'channelsOut': tf.FixedLenFeature([], tf.int64),
'data_raw': tf.FixedLenFeature([], tf.string),
'labels_raw': tf.FixedLenFeature([], tf.string)
})
labels = tf.decode_raw(featuresA['labels_raw'], tf.float32)
else:
featuresA = tf.parse_single_example(
value,
features={
'DataH': tf.FixedLenFeature([], tf.int64),
'DataW': tf.FixedLenFeature([], tf.int64),
'channelsIn': tf.FixedLenFeature([], tf.int64),
'data_raw': tf.FixedLenFeature([], tf.string)
})
feature = tf.decode_raw(featuresA['data_raw'], tf.float32)
print('setup_inputs')
print('Data H,W,#ch: %d,%d,%d -> Labels H,W,#ch %d,%d,%d' % (DataH,DataW,channelsIn,LabelsH,LabelsW,channelsOut))
print('------------------')
if myParams.myDict['InputMode'] == '1DFTy':
feature = tf.reshape(feature, [256, 256, 1])
feature = tf.random_crop(feature, [DataH, DataW, channelsIn])
mm=tf.reduce_mean(feature)
mx=tf.reduce_max(feature)
mx=tf.maximum(mx,1)
#feature = tf.Print(feature,[mm,mx],message='QQQ:')
#assert_op = tf.Assert(tf.greater(mx, 0), [mx])
#with tf.control_dependencies([assert_op]):
feature = tf.cast(feature/mx, tf.complex64)
Q=GT.TFGenerateRandomSinPhase(DataH, DataW)
IQ=feature*tf.reshape(Q,[DataH,DataW,channelsIn])
label=tf.concat([tf.real(IQ),tf.imag(IQ)],axis=2)
feature=label
HalfDataW=DataW/2
Id=np.hstack([np.arange(HalfDataW,DataW), np.arange(0,HalfDataW)])
Id=Id.astype(int)
IQ2=tf.reshape(IQ,IQ.shape[0:2])
feature=tf.fft(IQ2)
feature = tf.gather(feature,Id,axis=1)
feature = tf.reshape(feature, [DataH, DataW, channelsIn])
feature=tf.concat([tf.real(feature),tf.imag(feature)],axis=2)
features, labels = tf.train.batch([feature, label],
batch_size=batch_size,
num_threads=4,
capacity = capacity_factor*batch_size,
name='labels_and_features')
tf.train.start_queue_runners(sess=sess)
return features, labels
if myParams.myDict['InputMode'] == '1DFTx':
feature = tf.reshape(feature, [256, 256, 1])
feature = tf.random_crop(feature, [DataH, DataW, channelsIn])
mm=tf.reduce_mean(feature)
mx=tf.reduce_max(feature)
mx=tf.maximum(mx,1)
#feature = tf.Print(feature,[mm,mx],message='QQQ:')
#assert_op = tf.Assert(tf.greater(mx, 0), [mx])
#with tf.control_dependencies([assert_op]):
feature = tf.cast(feature/mx, tf.complex64)
Q=GT.TFGenerateRandomSinPhase(DataH, DataW)
IQ=feature*tf.reshape(Q,[DataH,DataW,channelsIn])
label=tf.concat([tf.real(IQ),tf.imag(IQ)],axis=2)
feature=label
HalfDataH=DataH/2
Id=np.hstack([np.arange(HalfDataH,DataH), np.arange(0,HalfDataH)])
Id=Id.astype(int)
IQ2=tf.reshape(IQ,IQ.shape[0:2])
IQ2 = tf.transpose(IQ2, perm=[1, 0])
feature=tf.fft(IQ2)
feature = tf.gather(feature,Id,axis=1)
feature = tf.transpose(feature, perm=[1,0])
feature = tf.reshape(feature, [DataH, DataW, channelsIn])
feature=tf.concat([tf.real(feature),tf.imag(feature)],axis=2)
features, labels = tf.train.batch([feature, label],
batch_size=batch_size,
num_threads=4,
capacity = capacity_factor*batch_size,
name='labels_and_features')
tf.train.start_queue_runners(sess=sess)
return features, labels
if myParams.myDict['InputMode'] == '2DFT':
feature = tf.reshape(feature, [256, 256, 1])
feature = tf.random_crop(feature, [DataH, DataW, channelsIn])
mm=tf.reduce_mean(feature)
mx=tf.reduce_max(feature)
mx=tf.maximum(mx,1)
#feature = tf.Print(feature,[mm,mx],message='QQQ:')
#assert_op = tf.Assert(tf.greater(mx, 0), [mx])
#with tf.control_dependencies([assert_op]):
feature = tf.cast(feature/mx, tf.complex64)
Q=GT.TFGenerateRandomSinPhase(DataH, DataW)
IQ=feature*tf.reshape(Q,[DataH,DataW,channelsIn])
label=tf.concat([tf.real(IQ),tf.imag(IQ)],axis=2)
feature=label
HalfDataH=DataH/2
HalfDataW=DataW/2
IdH=np.hstack([np.arange(HalfDataH,DataH), np.arange(0,HalfDataH)])
IdH=IdH.astype(int)
IdW=np.hstack([np.arange(HalfDataW,DataW), np.arange(0,HalfDataW)])
IdW=IdW.astype(int)
IQ2=tf.reshape(IQ,IQ.shape[0:2])
IQ2=tf.fft(IQ2)
IQ2=tf.gather(IQ2,IdW,axis=1)
IQ2 = tf.transpose(IQ2, perm=[1, 0])
feature=tf.fft(IQ2)
feature = tf.gather(feature,IdH,axis=1)
feature = tf.transpose(feature, perm=[1,0])
feature = tf.reshape(feature, [DataH, DataW, channelsIn])
feature=tf.concat([tf.real(feature),tf.imag(feature)],axis=2)
features, labels = tf.train.batch([feature, label],
batch_size=batch_size,
num_threads=4,
capacity = capacity_factor*batch_size,
name='labels_and_features')
tf.train.start_queue_runners(sess=sess)
return features, labels
if kKick:
filename_queue2 = tf.train.string_input_producer(filenames)
key2, value2 = reader.read(filename_queue2)
featuresA2 = tf.parse_single_example(
value2,
features={
'DataH': tf.FixedLenFeature([], tf.int64),
'DataW': tf.FixedLenFeature([], tf.int64),
'channelsIn': tf.FixedLenFeature([], tf.int64),
'data_raw': tf.FixedLenFeature([], tf.string)
})
feature2 = tf.decode_raw(featuresA2['data_raw'], tf.float32)
feature = tf.reshape(feature, [DataH, DataW, channelsIn])
feature2 = tf.reshape(feature2, [DataH, DataW, channelsIn])
feature.set_shape([None, None, channelsIn])
feature2.set_shape([None, None, channelsIn])
feature = tf.cast(feature, tf.float32)/tf.reduce_max(feature)
feature2 = tf.cast(feature2, tf.float32)/tf.reduce_max(feature)
feature= tf.concat([feature,feature*0,feature2,feature2*0], 2)
label=feature
features, labels = tf.train.batch([feature, label],
batch_size=batch_size,
num_threads=4,
capacity = capacity_factor*batch_size,
name='labels_and_features')
tf.train.start_queue_runners(sess=sess)
return features, labels
if myParams.myDict['InputMode'] == 'RegridTry1':
# FullData=scipy.io.loadmat('/media/a/f38a5baa-d293-4a00-9f21-ea97f318f647/home/a/TF/NMapIndTesta.mat')
FullData=scipy.io.loadmat(myParams.myDict['NMAP_FN'])
NMapCR=FullData['NMapCR']
NMapCR = tf.constant(NMapCR)
feature=tf.gather(feature,NMapCR,validate_indices=None,name=None)
feature = tf.reshape(feature, [DataH, DataW, channelsIn])
feature = tf.cast(feature, tf.float32)
labels = tf.reshape(labels, [LabelsH, LabelsW, channelsOut])
label = tf.cast(labels, tf.float32)
# Using asynchronous queues
features, labels = tf.train.batch([feature, label],
batch_size=batch_size,
num_threads=4,
capacity = capacity_factor*batch_size,
name='labels_and_features')
tf.train.start_queue_runners(sess=sess)
return features, labels
if myParams.myDict['InputMode'] == 'SMASHTry1':
feature = tf.reshape(feature, [DataH, DataW, channelsIn])
feature = tf.cast(feature, tf.float32)
labels = tf.reshape(labels, [LabelsH, LabelsW, channelsOut])
label = tf.cast(labels, tf.float32)
# Using asynchronous queues
features, labels = tf.train.batch([feature, label],
batch_size=batch_size,
num_threads=4,
capacity = capacity_factor*batch_size,
name='labels_and_features')
tf.train.start_queue_runners(sess=sess)
return features, labels
"""if myParams.myDict['Mode'] == 'RegridTry1C2':
FullData=scipy.io.loadmat('/media/a/f38a5baa-d293-4a00-9f21-ea97f318f647/home/a/TF/NMapIndC.mat')
NMapCR=FullData['NMapCRC']
NMapCR = tf.constant(NMapCR)
feature=tf.gather(feature,NMapCR,validate_indices=None,name=None)
feature = tf.reshape(feature, [DataH, DataW, channelsIn,2])
feature = tf.cast(feature, tf.float32)
labels = tf.reshape(labels, [LabelsH, LabelsW, channelsOut])
label = tf.cast(labels, tf.float32)
# Using asynchronous queues
features, labels = tf.train.batch([feature, label],
batch_size=batch_size,
num_threads=4,
capacity = capacity_factor*batch_size,
name='labels_and_features')
tf.train.start_queue_runners(sess=sess)
return features, labels"""
feature = tf.reshape(feature, [DataH, DataW, channelsIn])
labels = tf.reshape(labels, [LabelsH, LabelsW, channelsOut])
#print('44')
#example.ParseFromString(serialized_example)
#x_1 = np.array(example.features.feature['X'].float_list.value)
# Convert from [depth, height, width] to [height, width, depth].
#result.uint8image = tf.transpose(depth_major, [1, 2, 0])
feature.set_shape([None, None, channelsIn])
labels.set_shape([None, None, channelsOut])
# Crop and other random augmentations
#image = tf.image.random_flip_left_right(image)
#image = tf.image.random_saturation(image, .95, 1.05)
#image = tf.image.random_brightness(image, .05)
#image = tf.image.random_contrast(image, .95, 1.05)
#print('55')
#wiggle = 8
#off_x, off_y = 25-wiggle, 60-wiggle
#crop_size = 128
#crop_size_plus = crop_size + 2*wiggle
#print('56')
#image = tf.image.crop_to_bounding_box(image, off_y, off_x, crop_size_plus, crop_size_plus)
#print('57')
#image = tf.image.crop_to_bounding_box(image, 1, 2, crop_size, crop_size)
#image = tf.random_crop(image, [crop_size, crop_size, 3])
feature = tf.reshape(feature, [DataH, DataW, channelsIn])
feature = tf.cast(feature, tf.float32) #/255.0
labels = tf.reshape(labels, [LabelsH, LabelsW, channelsOut])
label = tf.cast(labels, tf.float32) #/255.0
#if crop_size != image_size:
# image = tf.image.resize_area(image, [image_size, image_size])
# The feature is simply a Kx downscaled version
#K = 1
#downsampled = tf.image.resize_area(image, [image_size//K, image_size//K])
#feature = tf.reshape(downsampled, [image_size//K, image_size//K, 3])
#feature = tf.reshape(downsampled, [image_size//K, image_size//K, 3])
#label = tf.reshape(image, [image_size, image_size, 3])
#feature = tf.reshape(image, [image_size, image_size, channelsIn])
#feature = tf.reshape(image, [1, image_size*image_size*2, channelsIn])
#label = tf.reshape(labels, [image_size, image_size, channelsOut])
# Using asynchronous queues
features, labels = tf.train.batch([feature, label],
batch_size=batch_size,
num_threads=4,
capacity = capacity_factor*batch_size,
name='labels_and_features')
tf.train.start_queue_runners(sess=sess)
return features, labels
def train_model(train_data):
td = train_data
summaries = tf.summary.merge_all()
RestoreSession = False
if not RestoreSession:
td.sess.run(tf.global_variables_initializer())
# lrval = FLAGS.learning_rate_start
learning_rate_start = myParams.myDict['learning_rate_start']
lrval = myParams.myDict['learning_rate_start']
start_time = time.time()
last_summary_time = time.time()
last_checkpoint_time = time.time()
done = False
batch = 0
print("lrval %f" % (lrval))
# assert FLAGS.learning_rate_half_life % 10 == 0
# Cache test features and labels (they are small)
test_feature, test_label = td.sess.run([td.test_features, td.test_labels])
# test_label = td.sess.run([td.test_features, td.test_labels])
G_LossV = np.zeros((1000000), dtype=np.float32)
filename = os.path.join(myParams.myDict['train_dir'], 'TrainSummary.mat')
feed_dictOut = {td.gene_minput: test_feature}
gene_output = td.sess.run(td.gene_moutput, feed_dict=feed_dictOut)
# _summarize_progress(td, test_label, gene_output, batch, 'out')
feed_dict = {td.learning_rate: lrval}
opsx = [td.gene_minimize, td.gene_loss]
_, gene_loss = td.sess.run(opsx, feed_dict=feed_dict)
# opsy = [td.gene_loss]
# gene_loss = td.sess.run(opsy, feed_dict=feed_dict)
# ops = [td.gene_minimize, td.disc_minimize, td.gene_loss, td.disc_real_loss, td.disc_fake_loss]
# _, _, gene_loss, disc_real_loss, disc_fake_loss = td.sess.run(ops, feed_dict=feed_dict)
batch += 1
# run_metadata = tf.RunMetadata()
gene_output = td.sess.run(td.gene_moutput, feed_dict=feed_dictOut)
# gene_output = td.sess.run(td.gene_moutput, options=tf.RunOptions(trace_level=tf.RunOptions.FULL_TRACE, output_partition_graphs=True), feed_dict=feed_dictOut,run_metadata=run_metadata)
# _summarize_progress(td, test_label, gene_output, batch, 'out')
# with open("/tmp/run2.txt", "w") as out:
# out.write(str(run_metadata))
# fetched_timeline = timeline.Timeline(run_metadata.step_stats)
# chrome_trace = fetched_timeline.generate_chrome_trace_format()
# with open('timeline_01.json', 'w') as f:
# f.write(chrome_trace)
# tl = timeline.Timeline(run_metadata.step_stats)
# print(tl.generate_chrome_trace_format(show_memory=True))
# trace_file = tf.gfile.Open(name='timeline', mode='w')
# trace_file.write(tl.generate_chrome_trace_format(show_memory=True))
feed_dict = {td.learning_rate: lrval}
# ops = [td.gene_minimize, td.disc_minimize, td.gene_loss, td.disc_real_loss, td.disc_fake_loss]
opsx = [td.gene_minimize, td.gene_loss]
_, gene_loss = td.sess.run(opsx, feed_dict=feed_dict)
batch += 1
gene_output = td.sess.run(td.gene_moutput, feed_dict=feed_dictOut)
# _summarize_progress(td, test_label, gene_output, batch, 'out')
# load model
#saver.restore(sess,tf.train.latest_checkpoint('./'))
# running model on data:test_feature
RunOnData = False
if RunOnData:
filenames = tf.gfile.ListDirectory('DataAfterpMat')
filenames = sorted(filenames)
#filenames = [os.path.join('DataAfterpMat', f) for f in filenames]
Ni = len(filenames)
OutBase = myParams.myDict['SessionName'] + '_OutMat'
tf.gfile.MakeDirs(OutBase)
#pdb.set_trace()
for index in range(Ni):
print(index)
print(filenames[index])
CurData = scipy.io.loadmat(
os.path.join('DataAfterpMat', filenames[index]))
Data = CurData['CurData']
Data = Data.reshape((1, 64, 64, 1))
test_feature = np.kron(np.ones((16, 1, 1, 1)), Data)
#test_feature = np.array(np.random.choice([0, 1], size=(16,64,64,1)), dtype='float32')
feed_dictOut = {td.gene_minput: test_feature}
gene_output = td.sess.run(td.gene_moutput, feed_dict=feed_dictOut)
filenameOut = os.path.join(OutBase,
filenames[index][:-4] + '_out.mat')
SOut = {}
SOut['X'] = gene_output[0]
scipy.io.savemat(filenameOut, SOut)
# pdb.set_trace()
#_summarize_progress(td, test_feature, test_label, gene_output, batch, 'out')
# to get value of var:
# ww=td.sess.run(td.gene_var_list[1])
if GT.getparam('ShowRealData') > 0:
# ifilename=os.path.join('RealData', 'b.mat')
if GT.getparam('InputMode') == 'RegridTry3F_B0T2S_ITS_MB':
MB = GT.getparam('MB')
nCh = GT.getparam('nccToUse')
nTSC = GT.getparam('nTimeSegments')
batch_size = myParams.myDict['batch_size']
channelsIn = myParams.myDict['channelsIn']
channelsOut = myParams.myDict['channelsOut']
LabelsH = myParams.myDict['LabelsH']
LabelsW = myParams.myDict['LabelsW']
H = LabelsH
W = LabelsW
TimePoints_ms = GT.getparam('TimePoints_ms')
SN = GT.getparam('SN') # H,W,f8
P = GT.getparam('P') # H,W,f8
nTraj = GT.getparam('nTraj')
cCAIPIVecZ = GT.getparam('cCAIPIVecZ') # MB,nTraj complex128
TSBF = GT.getparam('TSBF') # nTS,nTraj f64
Rec0FN = '/opt/data/RealDataMB/meas_MID426_gBP_2dSpiral_multiecho_ASL_2mm_iso_run1_FID34170_2Echos/Sli06_Rec0.mat'
B0TSFFN = '/opt/data/RealDataMB/meas_MID426_gBP_2dSpiral_multiecho_ASL_2mm_iso_run1_FID34170_2Echos/Sli06/B0TS.mat'
SensDataFN = '/opt/data/RealDataMB/meas_MID426_gBP_2dSpiral_multiecho_ASL_2mm_iso_run1_FID34170_2Echos/Sli06/SensCC1.mat'
RealDtFN = '/opt/data/RealDataMB/meas_MID426_gBP_2dSpiral_multiecho_ASL_2mm_iso_run1_FID34170_2Echos/Sli06/RealDataForNN.mat'
Rec0F = scipy.io.loadmat(Rec0FN)
Rec0 = Rec0F['Rec0'] # (96, 192) complex128
Rec0MB = np.transpose(np.reshape(Rec0, [H, MB, W]),
(0, 2, 1)) # H,W,MB complex128
B0TSF = scipy.io.loadmat(B0TSFFN)
CurB0 = B0TSF['CurB0'] # (96, 96, 2) f8, in Hz
# B0TSF.keys()
SensData = scipy.io.loadmat(SensDataFN)
# SensData.keys()
SensCCA = SensData['SensCCA'] # (96, 96, 16) complex128
SensCCB = SensData['SensCCB'] # (96, 96, 16) complex128
RealSensCCMB = np.stack((SensCCA, SensCCB),
axis=3) # (96, 96, 16,2) complex128
RealSensCCMB = RealSensCCMB[:, :, :nCh, :] # H,W,nCh,MB
RealData = scipy.io.loadmat(RealDtFN)
DataCC = RealData['DataCC'] # (10238, 16) (nTraj, ncc) complex64
DataCC = DataCC[:, :nCh] # (nTraj, nCh) complex64
CurB0_Hz = CurB0[:, :, np.newaxis, :]
RealTSC0 = np.complex64(
np.exp(1j * 2 * np.pi * CurB0_Hz *
GT.NP_addDim(TimePoints_ms) / 1000)) # (96, 96, 7, 2)
# Add T2* of 20
T2S_est_ms = 20
RealTSC0 = RealTSC0 * np.exp(
-GT.NP_addDim(TimePoints_ms) / T2S_est_ms)
WarmStart_ITS = Rec0MB[:, :,
np.newaxis, :] * RealTSC0 # H,W,nTS,MB
TSBFCAIPI = np.transpose(cCAIPIVecZ[:, :, np.newaxis],
(1, 2, 0)) * np.transpose(
TSBF[:, :, np.newaxis],
(1, 0, 2)) # nTraj,nTS,MB c128
cTSBFCAIPI = np.conj(
np.transpose(
TSBFCAIPI[:, :, :, np.newaxis, np.newaxis],
(0, 1, 3, 2, 4))) # nTraj,nTS,/nCh/,MB,/batch_size/
DataCCMB = DataCC[:, np.newaxis, :, np.newaxis,
np.newaxis] # nTraj,/nTS/,nCh,/MB/,/batch_size/
DataCCMB = DataCCMB * cTSBFCAIPI # nTraj,nTS,nCh,MB,/batch_size/
# Padded=sps_x_dense_vecs(np.conj(np.transpose(P)),DataCCMB)
Padded = np.conj(np.transpose(P)) * np.reshape(
DataCCMB, (nTraj, -1))
Padded = np.reshape(Padded, (H * 2, W * 2, nTSC, nCh, MB, 1))
Padded = np.transpose(Padded, (2, 3, 4, 5, 0, 1))
# Padded=np.reshape(Padded,(nTSC,nCh,MB,1,H*2,W*2))
IFPadded = np.fft.ifft2(Padded, axes=(-2, -1))
Cropped = IFPadded[:, :, :, :, :H, :
W] # nTS,nCh,MB,/batch_size/,H,W
CroppedP = np.transpose(
Cropped, (5, 4, 1, 2, 0, 3)) # H,W,nCh,MB,nTS,/batch_size/
Real_NUFFTHSig = np.sum(
CroppedP *
np.conj(RealSensCCMB[:, :, :, :, np.newaxis, np.newaxis]),
axis=2) # H,W,MB,nTS,/batch_size/
# RealSensCCMB=RealSensCCMB[:,:,:nCh,:] # H,W,nCh,MB
Real_NUFFTHSig = np.transpose(Real_NUFFTHSig, (4, 0, 1, 3, 2))
Real_NUFFTHSig = Real_NUFFTHSig * np.conj(
SN[np.newaxis, :, :, np.newaxis,
np.newaxis]) # /batch_size/,H,W,nTS,MB
RealSensCCMB1D = GT.NP_ConcatRIOn0(
np.reshape(RealSensCCMB, (-1, 1, 1)))
Real_feature = RealSensCCMB1D
RealTSC01D = GT.NP_ConcatRIOn0(np.reshape(RealTSC0, (-1, 1, 1)))
Real_feature = np.concatenate((Real_feature, RealTSC01D), axis=0)
WarmStart_ITS1D = GT.NP_ConcatRIOn0(
np.reshape(WarmStart_ITS, (-1, 1, 1)))
Real_feature = np.concatenate((Real_feature, WarmStart_ITS1D),
axis=0)
Real_NUFFTHSig1D = GT.NP_ConcatRIOn0(
np.reshape(Real_NUFFTHSig, (-1, 1, 1)))
Real_feature = np.concatenate((Real_feature, Real_NUFFTHSig1D),
axis=0)
Real_feature = np.tile(Real_feature, (batch_size, 1, 1, 1))
if myParams.myDict['InputMode'] == 'SPEN_FC':
ifilename = myParams.myDict['RealDataFN']
RealData = scipy.io.loadmat(ifilename)
RealData = RealData['Data']
# RealData=np.reshape(RealData,)
# RealData=RealData
Real_feature = RealData
if myParams.myDict['InputMode'] == 'SPEN_Local':
ifilename = myParams.myDict['RealDataFN']
RealData = scipy.io.loadmat(ifilename)
RealData = RealData['Data']
# RealData=RealData
Real_feature = RealData
if False:
if RealData.ndim == 2:
RealData = RealData.reshape(
(RealData.shape[0], RealData.shape[1], 1, 1))
if RealData.ndim == 3:
RealData = RealData.reshape(
(RealData.shape[0], RealData.shape[1], RealData.shape[2],
1))
if myParams.myDict['InputMode'] == 'RegridTry1' or myParams.myDict[
'InputMode'] == 'RegridTry2':
ifilename = myParams.myDict['RealDataFN']
RealData = scipy.io.loadmat(ifilename)
RealData = RealData['Data']
# FullData=scipy.io.loadmat('/media/a/f38a5baa-d293-4a00-9f21-ea97f318f647/home/a/TF/NMapIndTesta.mat')
FullData = scipy.io.loadmat(myParams.myDict['NMAP_FN'])
NMapCR = FullData['NMapCR']
batch_size = myParams.myDict['batch_size']
Real_feature = np.reshape(RealData[0], [RealData.shape[1]])
Real_feature = np.take(Real_feature, NMapCR)
Real_feature = np.tile(Real_feature, (batch_size, 1, 1, 1))
if myParams.myDict['InputMode'] == 'RegridTry3' or myParams.myDict[
'InputMode'] == 'RegridTry3M' or myParams.myDict[
'InputMode'] == 'RegridTry3F' or myParams.myDict[
'InputMode'] == 'RegridTry3FMB' or myParams.myDict[
'InputMode'] == 'RegridTry3FME':
ifilename = myParams.myDict['RealDataFN']
RealData = scipy.io.loadmat(ifilename)
RealData = RealData['Data']
batch_size = myParams.myDict['batch_size']
nTraj = myParams.myDict['nTraj']
RealDatancc = myParams.myDict['RealDatancc']
nccInData = myParams.myDict['nccInData']
# RealData=RealData
# RealData=np.reshape(RealData,[batch_size,RealDatancc,nTraj,2])
# RealData=RealData[:,0:nccInData,:,:]
# RealData=np.reshape(RealData,[batch_size,nTraj,RealDatancc,2])
# RealData=RealData[:,:,0:nccInData,:]
# RealData=np.reshape(RealData,[batch_size,-1])
RealData = RealData[0, :]
RealData = np.tile(RealData, (batch_size, 1))
Real_feature = np.reshape(
RealData, [RealData.shape[0], RealData.shape[1], 1, 1])
Real_dictOut = {td.gene_minput: Real_feature}
# LearningDecayFactor=np.power(2,(-1/FLAGS.learning_rate_half_life))
learning_rate_half_life = myParams.myDict['learning_rate_half_life']
LearningDecayFactor = np.power(2, (-1 / learning_rate_half_life))
# train_time=FLAGS.train_time
train_time = myParams.myDict['train_time']
QuickFailureTimeM = myParams.myDict['QuickFailureTimeM']
QuickFailureThresh = myParams.myDict['QuickFailureThresh']
summary_period = myParams.myDict['summary_period'] # in Minutes
checkpoint_period = myParams.myDict['checkpoint_period'] # in Minutes
DiscStartMinute = myParams.myDict['DiscStartMinute']
gene_output = td.sess.run(td.gene_moutput, feed_dict=feed_dictOut)
if myParams.myDict['ShowRealData'] > 0:
gene_RealOutput = td.sess.run(td.gene_moutput, feed_dict=Real_dictOut)
gene_output[0] = gene_RealOutput[0]
Asuffix = 'out_%06.4f' % (gene_loss)
_summarize_progress(td, test_label, gene_output, batch, Asuffix)
print("Adding to saver:")
var_listX = td.gene_var_list
var_listX = [v for v in var_listX if "Bank" not in v.name]
# var_listX = tf.sort(var_listX)
for line in var_listX:
print("Adding " + line.name + ' ' +
str(line.shape.as_list()))
print("Saver var list end")
saver = tf.train.Saver(var_listX)
# _save_checkpoint(td, batch,G_LossV,saver)
tf.get_default_graph().finalize()
while not done:
batch += 1
gene_loss = disc_real_loss = disc_fake_loss = -1.234
# elapsed = int(time.time() - start_time)/60
CurTime = time.time()
elapsed = (time.time() - start_time) / 60
# Update learning rate
lrval *= LearningDecayFactor
if (learning_rate_half_life < 1000): # in minutes
lrval = learning_rate_start * np.power(
0.5, elapsed / learning_rate_half_life)
#print("batch %d gene_l1_factor %f' " % (batch,FLAGS.gene_l1_factor))
# if batch==200:
if elapsed > DiscStartMinute:
FLAGS.gene_l1_factor = 0.9
RunDiscriminator = FLAGS.gene_l1_factor < 0.999
feed_dict = {td.learning_rate: lrval}
if RunDiscriminator:
ops = [
td.gene_minimize, td.disc_minimize, td.gene_loss,
td.disc_real_loss, td.disc_fake_loss
]
_, _, gene_loss, disc_real_loss, disc_fake_loss = td.sess.run(
ops, feed_dict=feed_dict)
else:
ops = [
td.gene_minimize, td.gene_loss, td.MoreOut, td.MoreOut2,
td.MoreOut3
]
_, gene_loss, MoreOut, MoreOut2, MoreOut3 = td.sess.run(
ops, feed_dict=feed_dict)
G_LossV[batch] = gene_loss
# ggg: Force phase only var
# VR = [v for v in tf.global_variables() if v.name == "gene/GEN_L004/add_Mult2DMCyCSharedOverFeat_weightR:0"][0]
# VI = [v for v in tf.global_variables() if v.name == "gene/GEN_L004/add_Mult2DMCyCSharedOverFeat_weightI:0"][0]
# VRX=td.sess.run(VR);
# VIX=td.sess.run(VI);
# VC=VRX+1J*VIX
# Norm=np.abs(VC)
# Norm[Norm == 0] = 0.00001
# VRX=VRX/Norm
# VIX=VIX/Norm
# VR.load(VRX, td.sess)
# VI.load(VIX, td.sess)
# VR = [v for v in tf.global_variables() if v.name == "gene/GEN_L005/add_Mult2DMCxCSharedOverFeat_weightR:0"][0]
# VI = [v for v in tf.global_variables() if v.name == "gene/GEN_L005/add_Mult2DMCxCSharedOverFeat_weightI:0"][0]
# VRX=td.sess.run(VR);
# VIX=td.sess.run(VI);
# VC=VRX+1J*VIX
# Norm=np.abs(VC)
# Norm[Norm == 0] = 0.00001
# VRX=VRX/Norm
# VIX=VIX/Norm
# VR.load(VRX, td.sess)
# VI.load(VIX, td.sess)
# VR = [v for v in tf.global_variables() if v.name == "gene/GEN_L004/einsum_weightR:0"][0]
# VI = [v for v in tf.global_variables() if v.name == "gene/GEN_L004/einsum_weightI:0"][0]
# VRX=td.sess.run(VR);
# VIX=td.sess.run(VI);
# HmngWnd=np.power(np.hamming(98),1)
# HmngWnd=np.reshape(HmngWnd,[98,1,1])
# VC=VRX +1j*VIX
# FVC=GT.gfft(VC,dim=0)
# FVC=FVC*HmngWnd
# VC=GT.gifft(FVC,dim=0)
# VYR=np.real(VC)
# VYI=np.imag(VC)
# VR.load(VYR, td.sess)
# VI.load(VYI, td.sess)
if batch % 10 == 0:
# pdb.set_trace()
# Show we are alive
#print('Progress[%3d%%], ETA[%4dm], Batch [%4d], G_Loss[%3.3f], D_Real_Loss[%3.3f], D_Fake_Loss[%3.3f]' %
# (int(100*elapsed/train_time), train_time - int(elapsed), batch, gene_loss, disc_real_loss, disc_fake_loss))
print(
'Progress[%3d%%], ETA[%4dm], Batch [%4d], G_Loss[%3.3f], D_Real_Loss[%3.3f], D_Fake_Loss[%3.3f], MoreOut[%3.3f, %3.3f, %3.3f]'
% (int(100 * elapsed / train_time), train_time - int(elapsed),
batch, gene_loss, disc_real_loss, disc_fake_loss, MoreOut,
MoreOut2, MoreOut3))
# VLen=td.gene_var_list.__len__()
# for i in range(0, VLen):
# print(td.gene_var_list[i].name);
# print(VRX.dtype)
# print(VRX)
# exit()
# var_23 = [v for v in tf.global_variables() if v.name == "gene/GEN_L020/C2D_weight:0"][0]
# tmp=td.sess.run(td.gene_var_list[i])
# v.load([2, 3], td.sess)
if np.isnan(gene_loss):
print('NAN!!')
done = True
# ggg: quick failure test
if elapsed > QuickFailureTimeM:
if gene_loss > QuickFailureThresh:
print('Quick failure!!')
done = True
else:
QuickFailureTimeM = 10000000
# Finished?
current_progress = elapsed / train_time
if current_progress >= 1.0:
done = True
StopFN = '/media/a/f38a5baa-d293-4a00-9f21-ea97f318f647/home/a/TF/stop.a'
if os.path.isfile(StopFN):
print('Stop file used!!')
done = True
try:
tf.gfile.Remove(StopFN)
except:
pass
# Update learning rate
# if batch % FLAGS.learning_rate_half_life == 0:
# lrval *= .5
# if batch % FLAGS.summary_period == 0:
if (CurTime - last_summary_time) / 60 > summary_period:
# Show progress with test features
# feed_dict = {td.gene_minput: test_feature}
gene_output = td.sess.run(td.gene_moutput, feed_dict=feed_dictOut)
if myParams.myDict['ShowRealData'] > 0:
gene_RealOutput = td.sess.run(td.gene_moutput,
feed_dict=Real_dictOut)
gene_output[0] = gene_RealOutput[0]
Asuffix = 'out_%06.4f' % (gene_loss)
_summarize_progress(td, test_label, gene_output, batch, Asuffix)
last_summary_time = time.time()
# if batch % FLAGS.checkpoint_period == 0:
SaveCheckpoint_ByTime = (CurTime -
last_checkpoint_time) / 60 > checkpoint_period
CheckpointFN = '/media/a/f38a5baa-d293-4a00-9f21-ea97f318f647/home/a/TF/save.a'
SaveCheckPointByFile = os.path.isfile(CheckpointFN)
if SaveCheckPointByFile:
tf.gfile.Remove(CheckpointFN)
if SaveCheckpoint_ByTime or SaveCheckPointByFile:
last_checkpoint_time = time.time()
# Save checkpoint
_save_checkpoint(td, batch, G_LossV, saver)
RunOnAllFN = '/media/a/f38a5baa-d293-4a00-9f21-ea97f318f647/home/a/TF/RunOnAll.a'
RunOnAllFNByFile = os.path.isfile(RunOnAllFN)
if RunOnAllFNByFile:
tf.gfile.Remove(RunOnAllFN)
for r in range(1, 81):
ifilenamePrefix = myParams.myDict['LoadAndRunOnData_Prefix']
# ifilename=ifilenamePrefix + f'{r:02}' + '.mat'
ifilename = ifilenamePrefix + '%02d.mat' % (r)
RealData = scipy.io.loadmat(ifilename)
RealData = RealData['Data']
if RealData.ndim == 2:
RealData = RealData.reshape(
(RealData.shape[0], RealData.shape[1], 1, 1))
if RealData.ndim == 3:
RealData = RealData.reshape(
(RealData.shape[0], RealData.shape[1],
RealData.shape[2], 1))
Real_feature = RealData
if myParams.myDict[
'InputMode'] == 'RegridTry1' or myParams.myDict[
'InputMode'] == 'RegridTry2':
# FullData=scipy.io.loadmat('/media/a/f38a5baa-d293-4a00-9f21-ea97f318f647/home/a/TF/NMapIndTesta.mat')
FullData = scipy.io.loadmat(myParams.myDict['NMAP_FN'])
NMapCR = FullData['NMapCR']
batch_size = myParams.myDict['batch_size']
Real_feature = np.reshape(RealData[0], [RealData.shape[1]])
Real_feature = np.take(Real_feature, NMapCR)
Real_feature = np.tile(Real_feature, (batch_size, 1, 1, 1))
Real_dictOut = {td.gene_minput: Real_feature}
gene_RealOutput = td.sess.run(td.gene_moutput,
feed_dict=Real_dictOut)
OnRealData = {}
OnRealDataM = gene_RealOutput
# filenamex = 'OnRealData' + f'{r:02}' + '.mat'
filenamex = 'OnRealData' + '%02d.mat' % (r)
LoadAndRunOnData_OutP = myParams.myDict[
'LoadAndRunOnData_OutP']
filename = os.path.join(LoadAndRunOnData_OutP, filenamex)
OnRealData['x'] = OnRealDataM
scipy.io.savemat(filename, OnRealData)
print('Saved recon of real data')
_save_checkpoint(td, batch, G_LossV, saver)
print('Finished training!')
def _train():
# LoadAndRunOnData=False
LoadAndRunOnData = myParams.myDict['LoadAndRunOnData'] > 0
if LoadAndRunOnData:
# Setup global tensorflow state
sess, summary_writer = setup_tensorflow()
# Prepare directories
# filenames = prepare_dirs(delete_train_dir=False)
# Setup async input queues
# features, labels = srez_input.setup_inputs(sess, filenames)
features, labels = srez_input.setup_inputs(sess, 1)
# Create and initialize model
[gene_minput, gene_moutput,
gene_output, gene_var_list,
disc_real_output, disc_fake_output, disc_var_list] = \
srez_modelBase.create_model(sess, features, labels)
# Restore variables from checkpoint
print("Adding to saver:")
var_listX = gene_var_list
var_listX = [v for v in var_listX if "Bank" not in v.name]
for line in var_listX:
print("Adding " + line.name + ' ' +
str(line.shape.as_list()))
print("Saver var list end")
saver = tf.train.Saver(var_listX)
# saver = tf.train.Saver()
filename = 'checkpoint_new'
# filename = os.path.join('/media/a/f38a5baa-d293-4a00-9f21-ea97f318f647/home/a/TF/srez/RegridTry1C2_TS2_dataNeighborhoodRCB0__2018-06-08_16-17-56_checkpoint', filename)
# filename = os.path.join('/media/a/f38a5baa-d293-4a00-9f21-ea97f318f647/home/a/TF/srez/RegridTry1C2_TS2_dataNeighborhoodRCB0__2018-06-09_19-44-17_checkpoint', filename)
# filename = os.path.join('/media/a/f38a5baa-d293-4a00-9f21-ea97f318f647/home/a/TF/srez/RegridTry1C2_TS__2018-06-29_10-39-13_checkpoint', filename)
checkpointP = myParams.myDict['LoadAndRunOnData_checkpointP']
filename = os.path.join(checkpointP, filename)
saver.restore(sess, filename)
if myParams.myDict['Mode'] == 'RegridTry1' or myParams.myDict[
'Mode'] == 'RegridTry1C' or myParams.myDict[
'Mode'] == 'RegridTry1C2' or myParams.myDict[
'Mode'] == 'RegridTry1C2_TS' or myParams.myDict[
'Mode'] == 'RegridTry1C2_TS2':
FullData = scipy.io.loadmat(myParams.myDict['NMAP_FN'])
NMapCR = FullData['NMapCR']
for r in range(1, myParams.myDict['HowManyToRun']):
# ifilename='/media/a/f38a5baa-d293-4a00-9f21-ea97f318f647/home/a/TF/srez/RealData/b_Ben14May_Sli5_r' + f'{r:02}' + '.mat'
# ifilename='/media/a/DATA/14May18/Ben/meas_MID109_gBP_VD11_U19_4min_FID17944/RealData/Sli11_r' + f'{r:02}' + '.mat'
if myParams.myDict['InputMode'] == 'Cart3SB':
# print('Loaded SensMaps, Shape %d %d %d %d' % (SensMapsSz[0],SensMapsSz[1],SensMapsSz[2],SensMapsSz[3]))
print('Cart3SB running on real data %d' % r)
# feature shape: (1048576, 1, 1) <dtype: 'float32'>
batch_size = myParams.myDict['batch_size']
# RealData=np.zeros((batch_size,1048576, 1, 1),np.float32)
# RealData=np.zeros((batch_size,640000, 1, 1),np.float32)
# Simulating RealData from ITS, Sens
MB = GT.getparam('MB')
TimePoints_ms = GT.getparam('TimePoints_ms')
nTSC = TimePoints_ms.shape[0]
nCh = GT.getparam('nccToUse')
LabelsH = myParams.myDict['LabelsH']
LabelsW = myParams.myDict['LabelsW']
H = LabelsH
W = LabelsW
SnsFN = '/opt/data/CCSensMaps.mat'
fS = h5py.File(SnsFN, 'r')
SensMaps = fS['SensCC']
SensMaps = SensMaps['real'] + 1j * SensMaps['imag']
SensMapsSz = SensMaps.shape
print('r Loaded SensMaps, Shape %d %d %d %d' %
(SensMapsSz[0], SensMapsSz[1], SensMapsSz[2],
SensMapsSz[3]))
SensMaps = SensMaps[:, :, :, :nCh]
NumSensMapsInFile = SensMaps.shape[0]
# IdxS=15
for b in range(0, MB):
# if b==1:
# IdxB2=tf.random_uniform([1],minval=12,maxval=19,dtype=tf.int32)
# IdxS=IdxS+IdxB2[0]
# IdxS=tf.cond(IdxS[0]>=NumSensMapsInFile, lambda: IdxS-NumSensMapsInFile, lambda: IdxS)
# Sens=np.squeeze(SensMaps[IdxS,:,:,:],axis=0)
Sens = (SensMaps[15, :, :, :])
Sens = Sens[:H, :W, :nCh]
# Sens = tf.image.random_flip_left_right(Sens)
# Sens = tf.image.random_flip_up_down(Sens)
# uS=tf.random_uniform([1])
# Sens=tf.cond(uS[0]<0.5, lambda: tf.identity(Sens), lambda: tf.image.rot90(Sens))
SensMsk = GT.NP_addDim(
np.sum(np.abs(Sens), axis=2) > 0).astype(np.complex64)
Sens = GT.NP_addDim(Sens)
if b == 0:
SensMB = Sens
SensMskMB = SensMsk
# else:
# SensMB=tf.concat([SensMB,Sens],axis=3) # SensMB H W nCh MB
# SensMskMB=tf.concat([SensMskMB,SensMsk],axis=2) # SensMskMB H W MB
# nToLoad=myParams.myDict['nToLoad']
# LoadAndRunOnData=myParams.myDict['LoadAndRunOnData']>0
# if LoadAndRunOnData:
nToLoad = 300
print('r loading images ' + time.strftime("%Y-%m-%d %H:%M:%S"))
GREBaseP = '/opt/data/'
SFN = GREBaseP + 'All_Orientation-0x.mat'
f = h5py.File(SFN, 'r')
I = f['CurSetAll'][0:nToLoad]
print('r Loaded images ' + time.strftime("%Y-%m-%d %H:%M:%S"))
SendTSCest = GT.getparam('SendTSCest') > 0
HamPow = GT.getparam('HamPow')
# def TFexpix(X): return tf.exp(tf.complex(tf.zeros_like(X),X))
def NPexpix(X):
return np.exp(1j * X)
for b in range(0, MB):
# TFI = tf.constant(np.int16(I))
# Idx=tf.random_uniform([1],minval=0,maxval=I.shape[0],dtype=tf.int32)
Idx = 133
Data4 = (I[Idx, :, :, :])
# Data4=tf.squeeze(tf.slice(I,[Idx[0],0,0,0],[1,-1,-1,-1]),axis=0)
# Data4 = tf.image.random_flip_left_right(Data4)
# Data4 = tf.image.random_flip_up_down(Data4)
# u1=tf.random_uniform([1])
# Data4=tf.cond(u1[0]<0.5, lambda: tf.identity(Data4), lambda: tf.image.rot90(Data4))
# Data4 = tf.random_crop(Data4, [H, W, 4])
# Data4 = tf.random_crop(Data4, [:H, :W, 4])
Data4 = Data4[:H, :W, :]
# M=tf.slice(Data4,[0,0,0],[-1,-1,1])
# Ph=tf.slice(Data4,[0,0,1],[-1,-1,1])
# feature=tf.cast(M,tf.complex64)*TFexpix(Ph)
M = Data4[:, :, 0]
Ph = Data4[:, :, 1]
feature = M.astype(np.complex64) * NPexpix(Ph)
feature = GT.NP_addDim(feature) * SensMskMB[:, :, b:b + 1]
T2S_ms = Data4[:, :, 2]
# T2S_ms = tf.where( T2S_ms<1.5, 10000 * tf.ones_like( T2S_ms ), T2S_ms )
T2S_ms[T2S_ms < 1.5] = 10000
B0_Hz = Data4[:, :, 3]
# B0_Hz=M*0
# T2S_ms = tf.where( tf.is_nan(T2S_ms), 10000 * tf.ones_like( T2S_ms ), T2S_ms )
T2S_ms[np.isnan(T2S_ms)] = 10000
# B0_Hz = tf.where( tf.is_nan(B0_Hz), tf.zeros_like( B0_Hz ), B0_Hz )
B0_Hz[np.isnan(B0_Hz)] = 0
if SendTSCest:
# HamPowA=10
HamPowA = HamPow
HamA = np.roll(np.hamming(H), np.int32(H / 2))
HamA = np.power(HamA, HamPowA)
HamXA = np.reshape(HamA, (1, H, 1))
HamYA = np.reshape(HamA, (1, 1, W))
B0_Hz_Smoothed = np.transpose(
GT.NP_addDim(B0_Hz.astype(np.complex64)),
(2, 0, 1))
B0_Hz_Smoothed = np.fft.fft2(B0_Hz_Smoothed)
B0_Hz_Smoothed = B0_Hz_Smoothed * HamXA
B0_Hz_Smoothed = B0_Hz_Smoothed * HamYA
B0_Hz_Smoothed = np.fft.ifft2(B0_Hz_Smoothed)
B0_Hz_Smoothed = np.transpose(B0_Hz_Smoothed,
(1, 2, 0))
B0_Hz_Smoothed = np.real(B0_Hz_Smoothed)
TSCest = np.exp(1j * 2 * np.pi *
(B0_Hz_Smoothed * TimePoints_ms /
1000).astype(np.complex64))
# TSCest=np.ones(TSCest.shape).astype(np.complex64)
print('TSCest shape: ' + str(TSCest.shape))
# TSCest=TSCest*0+1
# print('TSCest shape: ' + str(TSCest.shape))
# print('reducing B0')
# print('B0_Hz shape: ' + str(B0_Hz.shape))
# print('B0_Hz_Smoothed shape: ' + str(B0_Hz_Smoothed.shape))
# B0_Hz=B0_Hz-np.squeeze(B0_Hz_Smoothed)
# print('B0_Hz shape: ' + str(B0_Hz.shape))
# urand_ms=tf.random_uniform([1])*12
# urand_sec=(tf.random_uniform([1])*2-1)*3/1000
# feature=feature*tf.cast(tf.exp(-urand_ms/T2S_ms),tf.complex64)
# feature=feature*TFexpix(2*np.pi*B0_Hz*urand_sec)
mx = M.max()
mx = np.maximum(mx, 1)
mx = mx.astype(np.complex64)
feature = feature / mx
CurIWithPhase = feature
TSCM = np.exp(-TimePoints_ms / GT.NP_addDim(T2S_ms))
TSCP = np.exp(1j * 2 * np.pi *
(GT.NP_addDim(B0_Hz) * TimePoints_ms /
1000).astype(np.complex64))
TSC = TSCM.astype(np.complex64) * TSCP
ITSbase = CurIWithPhase * TSC # ITSbase is H,W,nTSC
TSC = GT.NP_addDim(TSC)
ITSbase = GT.NP_addDim(ITSbase)
if b == 0:
CurIWithPhaseMB = CurIWithPhase
TSCMB = TSC
ITSbaseMB = ITSbase
if SendTSCest:
TSCest = GT.NP_addDim(TSCest)
TSCMBest = TSCest
# else:
# CurIWithPhaseMB=tf.concat([CurIWithPhaseMB,CurIWithPhase],axis=2) # CurIWithPhaseMB H W MB
# TSCMB=tf.concat([TSCMB,TSC],axis=3) # TSCMB H W nTSC MB
# ITSbaseMB=tf.concat([ITSbaseMB,ITSbase],axis=3) # ITSbaseMB H W nTSC MB
# if SendTSCest:
# TSCMBest=tf.stack([TSCMBest,TSCest],axis=3)
print('r ok 2')
ITS_P = np.transpose(
GT.NP_addDim(ITSbaseMB),
(4, 0, 1, 2, 3)) # /batch_size/,H,W,nTSC,MB
Msk3 = np.zeros((H, W, nTSC, 1, 1, 1))
PEShifts = GT.getparam('PEShifts')
PEJump = GT.getparam('PEJump')
print('r Using PEShifts')
for i in range(nTSC):
Msk3[PEShifts[i]::PEJump, :, i, :, :, :] = 1
Msk3 = np.complex64(Msk3)
# GT.setparam('CartMask',Msk3)
Sens6 = SensMB[:, :, np.newaxis, :, :,
np.newaxis] # H,W,/nTS/,nCh,MB,/batch_size/
# AHA_ITS=GT.Cartesian_OPHOP_ITS_MB(ITS_P,Sens6,Msk3)
ITS = np.transpose(ITSbaseMB, (0, 3, 2, 1)) # H, nTSC, W
ITS = np.reshape(ITS, (H, W * nTSC * MB, 1))
ITS_RI = GT.NP_ConcatRIOn2(ITS)
Sensc = SensMB
Sens1D = GT.NP_ConcatRIOn0(np.reshape(Sensc, (-1, 1, 1)))
feature = Sens1D
AHA_ITS = GT.NP_Cartesian_OPHOP_ITS_MB(ITS_P, Sens6, Msk3)
# new simpler approach
if SendTSCest:
TSCMBest_P = np.transpose(
GT.NP_addDim(TSCMBest),
(4, 0, 1, 2, 3)) # /batch_size/,H,W,nTSC,MB
AHA_ITS = AHA_ITS * np.conj(TSCMBest_P)
# send AHA_ITS
AHA_ITS_1D = GT.NP_ConcatRIOn0(np.reshape(AHA_ITS, (-1, 1, 1)))
feature = np.concatenate((feature, AHA_ITS_1D), axis=0)
if SendTSCest:
TSCest1D = GT.NP_ConcatRIOn0(
np.reshape(TSCMBest_P, (-1, 1, 1)))
feature = np.concatenate((feature, TSCest1D), axis=0)
RealData = np.tile(feature, (batch_size, 1, 1, 1))
# End simulating RealData
Real_feature = RealData
else:
ifilenamePrefix = myParams.myDict['LoadAndRunOnData_Prefix']
# ifilename=ifilenamePrefix + f'{r:02}' + '.mat'
ifilename = ifilenamePrefix + '%02d.mat' % (r)
RealData = scipy.io.loadmat(ifilename)
RealData = RealData['Data']
if RealData.ndim == 2:
RealData = RealData.reshape(
(RealData.shape[0], RealData.shape[1], 1, 1))
if RealData.ndim == 3:
RealData = RealData.reshape(
(RealData.shape[0], RealData.shape[1],
RealData.shape[2], 1))
Real_feature = RealData
# if myParams.myDict['Mode'] == 'RegridTry1' or myParams.myDict['Mode'] == 'RegridTry1C' or myParams.myDict['Mode'] == 'RegridTry1C2' or myParams.myDict['Mode'] == 'RegridTry1C2_TS' or myParams.myDict['Mode'] == 'RegridTry1C2_TS2':
# batch_size=myParams.myDict['batch_size']
# Real_feature=np.reshape(RealData[0],[RealData.shape[1]])
# Real_feature=np.take(Real_feature,NMapCR)
# Real_feature=np.tile(Real_feature, (batch_size,1,1,1))
if myParams.myDict['InputMode'] == 'RegridTry1' or myParams.myDict[
'InputMode'] == 'RegridTry2':
# FullData=scipy.io.loadmat('/media/a/f38a5baa-d293-4a00-9f21-ea97f318f647/home/a/TF/NMapIndTesta.mat')
FullData = scipy.io.loadmat(myParams.myDict['NMAP_FN'])
NMapCR = FullData['NMapCR']
batch_size = myParams.myDict['batch_size']
Real_feature = np.reshape(RealData[0], [RealData.shape[1]])
Real_feature = np.take(Real_feature, NMapCR)
Real_feature = np.tile(Real_feature, (batch_size, 1, 1, 1))
Real_dictOut = {gene_minput: Real_feature}
gene_RealOutput = sess.run(gene_moutput, feed_dict=Real_dictOut)
OnRealData = {}
OnRealDataM = gene_RealOutput
# filenamex = 'OnRealData' + f'{r:02}' + '.mat'
filenamexBase = 'OnRealData' + '%02d' % (r)
filenamex = filenamexBase + '.mat'
LoadAndRunOnData_OutP = myParams.myDict['LoadAndRunOnData_OutP']
filename = os.path.join(LoadAndRunOnData_OutP, filenamex)
OnRealData['x'] = OnRealDataM
scipy.io.savemat(filename, OnRealData)
image = np.sqrt(
np.square(OnRealDataM[0, -H:, :(W * 3), 0]) +
np.square(OnRealDataM[0, -H:, :(W * 3), 1]))
filenamep = filenamexBase + '.png'
filename = os.path.join(LoadAndRunOnData_OutP, filenamep)
imageio.imwrite(filename, image)
print('Saved recon of real data')
exit()
# Setup global tensorflow state
sess, summary_writer = setup_tensorflow()
# Prepare directories
all_filenames = prepare_dirs(delete_train_dir=True)
# Separate training and test sets
#train_filenames = all_filenames[:-FLAGS.test_vectors]
train_filenames = all_filenames
#test_filenames = all_filenames[-FLAGS.test_vectors:]
# TBD: Maybe download dataset here
#pdb.set_trace()
# ggg Signal Bank stuff:
if myParams.myDict['BankSize'] > 0:
if myParams.myDict['InputMode'] == 'RegridTry3FMB':
BankSize = myParams.myDict['BankSize'] * 2
# BankInit=np.zeros([BankSize,myParams.myDict['DataH'],1,1])
# LBankInit=np.zeros([BankSize,myParams.myDict['LabelsH'],myParams.myDict['LabelsW'], 2])
with tf.variable_scope("aaa"):
Bank = tf.get_variable(
"Bank",
shape=[BankSize, myParams.myDict['DataH'], 1, 1],
dtype=tf.float32,
trainable=False)
LBank = tf.get_variable("LBank",
shape=[
BankSize,
myParams.myDict['LabelsH'],
myParams.myDict['LabelsW'], 2
],
dtype=tf.float32,
trainable=False)
# LBank=tf.get_variable("LBank",initializer=tf.cast(LBankInit, tf.float32),dtype=tf.float32,trainable=False)
else:
BankSize = myParams.myDict['BankSize']
BankInit = np.zeros([BankSize, myParams.myDict['DataH'], 1, 1])
LBankInit = np.zeros([
BankSize, myParams.myDict['LabelsH'],
myParams.myDict['LabelsW'], 2
])
with tf.variable_scope("aaa"):
# Bank=tf.get_variable("Bank",initializer=tf.cast(BankInit, tf.float32),dtype=tf.float32)
Bank = tf.get_variable(
"Bank",
shape=[BankSize, myParams.myDict['DataH'], 1, 1],
dtype=tf.float32,
trainable=False)
LBank = tf.get_variable("LBank",
shape=[
BankSize,
myParams.myDict['LabelsH'],
myParams.myDict['LabelsW'], 2
],
dtype=tf.float32,
trainable=False)
# LBank=tf.get_variable("LBank",initializer=tf.cast(LBankInit, tf.float32),dtype=tf.float32)
init_new_vars_op = tf.variables_initializer([Bank, LBank])
sess.run(init_new_vars_op)
# ggg end Signal Bank stuff:
# Setup async input queues
train_features, train_labels = srez_input.setup_inputs(
sess, train_filenames)
# test_features, test_labels = srez_input.setup_inputs(sess, train_filenames,TestStuff=True)
test_features = train_features
test_labels = train_labels
#test_features, test_labels = srez_input.setup_inputs(sess, test_filenames)
print('starting' + time.strftime("%Y-%m-%d %H:%M:%S"))
print('train_features %s' % (train_features))
print('train_labels %s' % (train_labels))
# Add some noise during training (think denoising autoencoders)
noise_level = myParams.myDict['noise_level']
AddNoise = noise_level > 0.0
if AddNoise:
noisy_train_features = train_features + tf.random_normal(
train_features.get_shape(), stddev=noise_level)
else:
noisy_train_features = train_features
# Create and initialize model
[gene_minput, gene_moutput,
gene_output, gene_var_list,
disc_real_output, disc_fake_output, disc_var_list] = \
srez_modelBase.create_model(sess, noisy_train_features, train_labels)
# gene_VarNamesL=[];
# for line in gene_var_list: gene_VarNamesL.append(line.name+' ' + str(line.shape.as_list()))
# gene_VarNamesL.sort()
# for line in gene_VarNamesL: print(line)
# # var_23 = [v for v in tf.global_variables() if v.name == "gene/GEN_L020/C2D_weight:0"][0]
# for line in sess.graph.get_operations(): print(line)
# Gen3_ops=[]
# for line in sess.graph.get_operations():
# if 'GEN_L003' in line.name:
# Gen3_ops.append(line)
# # LL=QQQ.outputs[0]
# for x in Gen3_ops: print(x.name +' ' + str(x.outputs[0].shape))
# GenC2D_ops= [v for v in sess.graph.get_operations()]
# GenC2D_ops= [v for v in tf.get_operations() if "weight" in v.name]
# GenC2D_ops= [v for v in GenC2D_ops if "C2D" in v.name]
# for x in GenC2D_ops: print(x.name +' ' + str(x.outputs[0].shape))
# for x in GenC2D_ops: print(x.name)
AEops = [
v for v in sess.graph.get_operations()
if "AE" in v.name and not ("_1/" in v.name)
]
# AEops = [v for v in td.sess.graph.get_operations() if "Pixel" in v.name and not ("_1/" in v.name) and not ("opti" in v.name) and not ("Assign" in v.name) and not ("read" in v.name) and not ("Adam" in v.name)]
AEouts = [v.outputs[0] for v in AEops]
varsForL1 = AEouts
# varsForL1=AEouts[0:-1]
# varsForL1=AEouts[1:]
# for line in sess.graph.get_operations():
# if 'GEN_L003' in line.name:
# Gen3_ops.append(line)
# # LL=QQQ.outputs[0]
# for x in Gen3_ops: print(x.name +' ' + str(x.outputs[0].shape))
print("Vars for l2 loss:")
varws = [
v for v in tf.global_variables()
if (("weight" in v.name) or ("ConvNet" in v.name))
]
varsForL2 = [v for v in varws if "C2D" in v.name]
varsForL2 = [v for v in varws if "disc" not in v.name]
varsForL2 = [v for v in varws if "bias" not in v.name]
for line in varsForL2:
print(line.name + ' ' + str(line.shape.as_list()))
print("Vars for Phase-only loss:")
varws = [v for v in tf.global_variables() if "weight" in v.name]
varsForPhaseOnly = [v for v in varws if "SharedOverFeat" in v.name]
for line in varsForPhaseOnly:
print(line.name + ' ' + str(line.shape.as_list()))
# pdb.set_trace()
gene_loss, MoreOut, MoreOut2, MoreOut3 = srez_modelBase.create_generator_loss(
disc_fake_output, gene_output, train_features, train_labels, varsForL1,
varsForL2, varsForPhaseOnly)
disc_real_loss, disc_fake_loss = \
srez_modelBase.create_discriminator_loss(disc_real_output, disc_fake_output)
disc_loss = tf.add(disc_real_loss, disc_fake_loss, name='disc_loss')
(global_step, learning_rate, gene_minimize, disc_minimize) = \
srez_modelBase.create_optimizers(gene_loss, gene_var_list, disc_loss, disc_var_list)
# Train model
train_data = TrainData(locals())
#pdb.set_trace()
# ggg: to restore session
RestoreSession = False
if RestoreSession:
saver = tf.train.Saver()
filename = 'checkpoint_new'
filename = os.path.join(myParams.myDict['checkpoint_dir'], filename)
saver.restore(sess, filename)
srez_train.train_model(train_data)
import os import sys # sys.path.insert(0, '/media/a/H2/home/a/TF/') sys.path.insert(0, '/opt/data/TF/') import time import GTools as GT # os.environ['TF_CPP_MIN_LOG_LEVEL'] = '0' # os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2' os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3' # HomeA='/home/deni/' HomeA = GT.getHome() sys.path.insert(0, HomeA + 'TF/') # DatasetsBase='/home/deni/' DatasetsBase = GT.getDatasetsBase() os.chdir(HomeA + 'TF/srezN') # os.environ['CUDA_VISIBLE_DEVICES'] = '1' # os.environ['CUDA_VISIBLE_DEVICES'] = '' import pdb # import srez_demo import srez_input import srez_model
import time import GTools as GT # os.environ['TF_CPP_MIN_LOG_LEVEL'] = '0' # os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2' os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3' # HomeA=GT.getHome() HomeA = os.getcwd() # os.chdir(HomeA + 'TF/srez') ParamFN = HomeA + '/Params.txt' # DatasetsBase='/home/deni/' DatasetsBase = GT.getDatasetsBase() os.environ['CUDA_VISIBLE_DEVICES'] = '0' # os.environ['CUDA_VISIBLE_DEVICES'] = '' import pdb # import srez_demo import srez_input import srez_model import srez_train import srez_modelBase import os.path import random import numpy as np
def _generator_model(sess, features, labels, channels):
# Upside-down all-convolutional resnet
mapsize = 3
mapsize = myParams.myDict['MapSize']
res_units = [256, 128, 96]
old_vars = tf.global_variables()
# See Arxiv 1603.05027
model = Model('GEN', features)
# H=FLAGS.LabelsH;
# W=FLAGS.LabelsW;
H = myParams.myDict['LabelsH']
W = myParams.myDict['LabelsW']
channelsOut = myParams.myDict['channelsOut']
batch_size = myParams.myDict['batch_size']
DataH = myParams.myDict['DataH']
print("_generator_model")
print("%d %d %d" % (H, W, channels))
if myParams.myDict['NetMode'] == 'SPEN_Local':
print("SPEN_Local mode")
model.add_Split4thDim(2) # now (16, H, W, HNeighbors, 2)
model.add_PixelwiseMultC(1) #,NamePrefix='MapsForMat')
model.remove_5thDim()
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == 'SPEN_FC':
print("SPEN_FC mode")
model.add_5thDim()
model.add_Permute45()
model.add_Mult2DMCxC(H, 1)
model.remove_5thDim()
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == 'SMASH1DFTxyC_YCC':
print("SMASH1DFTxyC_YCC mode")
# model.print_size('AAA') # (16, 64, 128, 16)
model.add_Split4thDim(2) # now (16, 64, 128, 8, 2)
DFTM = GT.DFT_matrix(H)
IDFTM = GT.IDFT_matrix(H)
DFTM_Half = GT.DFT_matrix(64)
IDFTM_Half = GT.IDFT_matrix(64)
# back to image space on RO
model.add_Mult2DMCyCSharedOverFeat(W, 1, Trainable=False, InitC=IDFTM)
# YCC: also on PE
model.add_Mult2DMCxCSharedOverFeat(64,
1,
Trainable=False,
InitC=IDFTM_Half)
# CC part
ncc = myParams.myDict['CC_channels']
# CC: model.add_conv2dC(ncc,mapsize=1) # now (16, 64, 128, ncc, 2)
model.add_einsumC('abcd,bcdx->abcx', [64, 128, 8, ncc])
# back to k-space space on RO
model.add_Mult2DMCyCSharedOverFeat(W, 1, Trainable=False, InitC=DFTM)
# YCC: also on PE
model.add_Mult2DMCxCSharedOverFeat(64,
1,
Trainable=False,
InitC=DFTM_Half)
# now conv, from 8 to 2
model.add_conv2dC(2, mapsize=3) # now (16, 64, 128, 2, 2)
# now put combine the 2 with the 64
model.add_Permute([0, 1, 3, 2, 4]) # now (16, 64, 2, 128, 2)
model.add_Reshape([16, 128, 128, 1, 2])
# model.add_Mult2DMCxCSharedOverFeat(H,1,NamePrefix='MapsForMat')
model.add_Mult2DMCxCSharedOverFeat(H, 1, Trainable=False, InitC=IDFTM)
model.add_Mult2DMCyCSharedOverFeat(W, 1, Trainable=False, InitC=IDFTM)
model.remove_5thDim()
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == 'SMASH1DFTxyC_XCC':
print("SMASH1DFTxyC_XCC mode")
# model.print_size('AAA') # (16, 64, 128, 16)
model.add_Split4thDim(2) # now (16, 64, 128, 8, 2)
DFTM = GT.DFT_matrix(H)
IDFTM = GT.IDFT_matrix(H)
# back to image space on RO
model.add_Mult2DMCyCSharedOverFeat(W, 1, Trainable=False, InitC=IDFTM)
# CC part
ncc = myParams.myDict['CC_channels']
# CC: model.add_conv2dC(ncc,mapsize=1) # now (16, 64, 128, ncc, 2)
model.add_einsumC('abcd,bcdx->abcx', [64, 128, 8, ncc])
# back to k-space space on RO
model.add_Mult2DMCyCSharedOverFeat(W, 1, Trainable=False, InitC=DFTM)
# now conv, from 8 to 2
model.add_conv2dC(2, mapsize=3) # now (16, 64, 128, 2, 2)
# now put combine the 2 with the 64
model.add_Permute([0, 1, 3, 2, 4]) # now (16, 64, 2, 128, 2)
model.add_Reshape([16, 128, 128, 1, 2])
# model.add_Mult2DMCxCSharedOverFeat(H,1,NamePrefix='MapsForMat')
model.add_Mult2DMCxCSharedOverFeat(H, 1, Trainable=False, InitC=IDFTM)
model.add_Mult2DMCyCSharedOverFeat(W, 1, Trainable=False, InitC=IDFTM)
model.remove_5thDim()
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == 'SMASH1DFTxyC_GCC':
print("SMASH1DFTxyC_GCC mode")
# model.print_size('AAA') # (16, 64, 128, 16)
model.add_Split4thDim(2) # now (16, 64, 128, 8, 2)
DFTM = GT.DFT_matrix(H)
IDFTM = GT.IDFT_matrix(H)
# back to image space on RO
model.add_Mult2DMCyCSharedOverFeat(W, 1, Trainable=False, InitC=IDFTM)
# CC part
ncc = myParams.myDict['CC_channels']
# CC: model.add_conv2dC(ncc,mapsize=1) # now (16, 64, 128, ncc, 2)
model.add_einsumC('abcd,cdx->abcx', [128, 8, ncc])
# back to k-space space on RO
model.add_Mult2DMCyCSharedOverFeat(W, 1, Trainable=False, InitC=DFTM)
# now conv, from 8 to 2
model.add_conv2dC(2, mapsize=3) # now (16, 64, 128, 2, 2)
# now put combine the 2 with the 64
model.add_Permute([0, 1, 3, 2, 4]) # now (16, 64, 2, 128, 2)
model.add_Reshape([16, 128, 128, 1, 2])
# model.add_Mult2DMCxCSharedOverFeat(H,1,NamePrefix='MapsForMat')
model.add_Mult2DMCxCSharedOverFeat(H, 1, Trainable=False, InitC=IDFTM)
model.add_Mult2DMCyCSharedOverFeat(W, 1, Trainable=False, InitC=IDFTM)
model.remove_5thDim()
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == 'SMASH1DFTxyC_SCC':
print("SMASH1DFTxyC_SCC mode")
# model.print_size('AAA') # (16, 64, 128, 16)
model.add_Split4thDim(2) # now (16, 64, 128, 8, 2)
DFTM = GT.DFT_matrix(H)
IDFTM = GT.IDFT_matrix(H)
# back to image space on RO
model.add_Mult2DMCyCSharedOverFeat(W, 1, Trainable=False, InitC=IDFTM)
# CC part
ncc = myParams.myDict['CC_channels']
model.add_conv2dC(ncc, mapsize=1) # now (16, 64, 128, ncc, 2)
# back to k-space space on RO
model.add_Mult2DMCyCSharedOverFeat(W, 1, Trainable=False, InitC=DFTM)
# now conv, from 8 to 2
model.add_conv2dC(2, mapsize=3) # now (16, 64, 128, 2, 2)
# now put combine the 2 with the 64
model.add_Permute([0, 1, 3, 2, 4]) # now (16, 64, 2, 128, 2)
model.add_Reshape([16, 128, 128, 1, 2])
# model.add_Mult2DMCxCSharedOverFeat(H,1,NamePrefix='MapsForMat')
model.add_Mult2DMCxCSharedOverFeat(H, 1, Trainable=False, InitC=IDFTM)
model.add_Mult2DMCyCSharedOverFeat(W, 1, Trainable=False, InitC=IDFTM)
model.remove_5thDim()
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == 'SMASH1DFTxyC':
print("1DFTxyCMaps mode")
# model.print_size('AAA') # (16, 64, 128, 16)
model.add_Split4thDim(2) # now (16, 64, 128, 8, 2)
# now conv, from 8 to 2
model.add_conv2dC(2, mapsize=3) # now (16, 64, 128, 2, 2)
# now put combine the 2 with the 64
model.add_Permute([0, 1, 3, 2, 4]) # now (16, 64, 2, 128, 2)
model.add_Reshape([16, 128, 128, 1, 2])
IDFTM = GT.IDFT_matrix(H)
# model.add_Mult2DMCxCSharedOverFeat(H,1,NamePrefix='MapsForMat')
model.add_Mult2DMCxCSharedOverFeat(H, 1, Trainable=False, InitC=IDFTM)
model.add_Mult2DMCyCSharedOverFeat(W, 1, Trainable=False, InitC=IDFTM)
model.remove_5thDim()
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == '1DFTxyCMaps':
print("1DFTxyCMaps mode")
# model.print_size('AAA') # (16, 128, 128, 16)
model.add_Split4thDim(2) # now (16, 128, 128, 8, 2)
# model.print_size('CCC')
# model.add_Permute45()
model.add_Mult2DMCxCSharedOverFeat(H, 1, NamePrefix='MapsForMat')
model.add_Mult2DMCyCSharedOverFeat(W, 1)
model.add_PixelwiseMultC(1) #,NamePrefix='MapsForMat')
model.remove_5thDim()
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == '2DFTC':
print("2DFTC mode")
model.add_5thDim()
model.add_Permute45()
model.add_Mult2DMCxC(H * W, 1)
model.remove_5thDim()
model.add_reshapeTo4D(H, W)
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == '1DFTxyC':
print("1DFTxyC mode")
model.add_5thDim()
model.add_Permute45()
model.add_Mult2DMCxC(H, 1)
model.add_Mult2DMCyC(W, 1)
model.remove_5thDim()
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == '1DFTxC':
print("1DFTxC mode")
model.add_5thDim()
model.add_Permute45()
model.add_Mult2DMCxC(H, 1)
model.remove_5thDim()
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == '1DFTyC':
print("1DFTyC mode")
model.add_5thDim()
model.add_Permute45()
model.add_Mult2DMCyC(W, 1)
model.remove_5thDim()
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == '1DFTy':
print("1DFTy mode")
model.add_Mult2DMCy(W, channelsOut)
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == '1DFTx':
print("1DFTx mode")
model.add_Mult2DMCx(H, channelsOut)
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == '2DFT':
print("2DFT mode")
model.add_Mult2DMCy(W, channelsOut)
model.add_Mult2DMCx(H, channelsOut)
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
# if myParams.myDict['NetMode'] == 'RegridTry1':
# print("RegridTry1 mode")
# model.add_PixelwiseMult(2, stddev_factor=1.0)
# model.add_Mult2DMCy(W,channelsOut)
# model.add_Mult2DMCx(H,channelsOut)
# new_vars = tf.global_variables()
# gene_vars = list(set(new_vars) - set(old_vars))
# return model.get_output(), gene_vars
# if myParams.myDict['NetMode'] == 'RegridTry1C':
# print("RegridTry1C mode")
# addBias=myParams.myDict['CmplxBias']>0
# if addBias:
# print("with bias")
# else:
# print("without bias")
# model.add_PixelwiseMult(2, stddev_factor=1.0)
# model.add_5thDim()
# model.add_Permute45()
# model.add_Mult2DMCyC(W,1,add_bias=addBias)
# model.add_Mult2DMCxC(H,1,add_bias=addBias)
# model.remove_5thDim()
# new_vars = tf.global_variables()
# gene_vars = list(set(new_vars) - set(old_vars))
# return model.get_output(), gene_vars
# if myParams.myDict['NetMode'] == 'RegridTry1C2':
# print("RegridTry1C2 mode")
# addBias=myParams.myDict['CmplxBias']>0
# if addBias:
# print("with bias")
# else:
# print("without bias")
# model.add_Split4thDim(2)
# model.add_PixelwiseMultC(1, stddev_factor=1.0)
# model.add_Mult2DMCyC(W,1,add_bias=addBias)
# model.add_Mult2DMCxC(H,1,add_bias=addBias)
# model.remove_5thDim()
# new_vars = tf.global_variables()
# gene_vars = list(set(new_vars) - set(old_vars))
# return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == 'RegridTry3C2_TS_WithTSB':
print("RegridTry3C2_TS_WithTSB mode")
FullData = scipy.io.loadmat(myParams.myDict['NMAP_FN'])
NMapCR = FullData['NMapCR']
NMapCR = tf.constant(NMapCR)
aDataH = myParams.myDict['aDataH']
aDataW = myParams.myDict['aDataW']
achannelsIn = myParams.myDict['achannelsIn']
nTS = myParams.myDict['nTimeSegments']
nccInData = myParams.myDict['nccInData']
nTraj = myParams.myDict['nTraj']
HalfDataH = np.int32(DataH / 2)
# 133068/2 = 66534
# model.print_shape('Start') # now 16,133068,1,1
# model.add_Permute([0,2,3,1])
# model.add_Split4thDim(2) # now 16,1,1,133068/2,2C
model.add_Reshape([batch_size, 1, 1, HalfDataH, 2])
# Now do TSB
model.add_Permute([0, 3, 2, 1, 4]) # now 16,133068/2,1,1,2C
model.add_Reshape([batch_size, nTraj, nccInData, 1, 2])
model.add_Permute([0, 2, 1, 3, 4]) # now 16 13 5118 1 2
model.add_Reshape([batch_size * nccInData, 1, nTraj, 1, 2])
model.add_PixelwiseMultC(
nTS, stddev_factor=1.0) # This is TSB. After: 16*13,1,5118,nTS,2
model.add_Reshape([batch_size, nccInData, nTraj, nTS, 2])
model.add_Permute([2, 1, 4, 0, 3]) # now 5118 13 2 16 nTS
model.add_Reshape([nTraj * nccInData * 2, batch_size * nTS, 1, 1])
# model.add_Permute([1,0,2,3])
# model.print_shape()
feature = model.get_output()
feature = tf.gather(feature, NMapCR, validate_indices=None, name=None)
# feature = tf.reshape(feature, [aDataH, aDataW, achannelsIn])
model.add_PutInOutput(feature) # After we're 131,131,192,16*nTS
model.add_Permute([3, 0, 1, 2, 4, 5]) # Now 16*nTS,131,131,192,1,1
# model.add_Reshape([batch_size,nTS,aDataH,aDataW,2,96])
# model.print_shape()
model.add_Reshape([batch_size * nTS, aDataH, aDataW,
achannelsIn]) # Now 16*nTS,131,131,192
UseSharedWightesInRelaxedFT = myParams.myDict[
'UseSharedWightesInRelaxedFT'] > 0
addBias = myParams.myDict['CmplxBias'] > 0
if addBias:
print("with bias")
else:
print("without bias")
model.add_Split4thDim(
2) # Now we're batch_size*nTS, kH,kW, Neighbors(12)*Channels(8),2C
model.add_PixelwiseMultC(
1, stddev_factor=1.0) # After we're batch_size*nTS,kH,kW,1,2C
# AfterRegrid_ForOut = tf.identity(model.get_output(), name="AfterRegrid_ForOut")
# model.add_PutInOutput(AfterRegrid_ForOut)
model.add_Reshape([batch_size, nTS, aDataH, aDataW, 2])
model.add_Permute([0, 2, 3, 1,
4]) # After we're batch_size,kH,kW,nTS, 2C
# AfterRegridP_ForOut = tf.identity(model.get_output(), name="AfterRegridP_ForOut")
# model.add_PutInOutput(AfterRegridP_ForOut)
# Now continuing as with no TSB
MM = GT.gDFT_matrix(np.linspace(-50, 50, aDataH), H)
MM = np.transpose(MM, axes=[1, 0])
if UseSharedWightesInRelaxedFT:
model.add_Mult2DMCyCSharedOverFeat(W,
1,
add_bias=addBias,
Trainable=False,
InitC=MM,
NamePrefix='FTy')
model.add_Mult2DMCxCSharedOverFeat(H,
1,
add_bias=addBias,
Trainable=False,
InitC=MM,
NamePrefix='FTx')
else:
model.add_Mult2DMCyC(W, 1, add_bias=addBias)
model.add_Mult2DMCxC(H, 1, add_bias=addBias)
# AfterFT_ForOut = tf.identity(model.get_output(), name="AfterFT_ForOut")
# model.add_PutInOutput(AfterFT_ForOut)
# now supposedly batch_size,H,W,nTS
model.add_PixelwiseMultC(
1, stddev_factor=1.0, NamePrefix='TSC'
) # This collecting the different TS to the final image.
# AfterTSC_ForOut = tf.identity(model.get_output(), name="AfterTSC_ForOut")
# model.add_PutInOutput(AfterTSC_ForOut)
model.remove_5thDim()
# EndForOut = tf.identity(model.get_output(), name="EndForOut")
# model.add_PutInOutput(EndForOut)
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == 'RegridTry3C2_TS':
print("RegridTry3C2_TS mode")
aDataH = myParams.myDict['aDataH']
kMax = myParams.myDict['kMax']
aDataW = myParams.myDict['aDataW']
# achannelsIn=myParams.myDict['achannelsIn']
nTS = myParams.myDict['nTimeSegments']
UseSharedWightesInRelaxedFT = myParams.myDict[
'UseSharedWightesInRelaxedFT'] > 0
RelaxedFT = myParams.myDict['RelaxedFT'] > 0
addBias = myParams.myDict['CmplxBias'] > 0
# FullData=scipy.io.loadmat(myParams.myDict['NMAP_FN'])
# NMapCR=FullData['NMapCR']
# NMapCR = tf.constant(NMapCR)
nccInData = myParams.myDict['nccInData']
# ncc=8
ncc = myParams.myDict['nccToUse']
nNeighbors = myParams.myDict['nNeighbors']
achannelsIn = ncc * nNeighbors * 2
# T=scipy.io.loadmat('/media/a/DATA/180628_AK/meas_MID244_gBP_VD11_U19_G35S155_4min_FID22439/Traj.mat')
# Traj=T['Traj'][0:2,:]
# BaseNUFTDataP='/media/a/DATA/13May18/Me/meas_MID409_gBP_VD11_U19_7ADCs_FID17798/'
# BaseNUFTDataP='/media/a/DATA/11Jul18/RL/meas_MID149_gBP_VD11_U19_G35S155_FID23846/'
BaseNUFTDataP = myParams.myDict['BaseNUFTDataP']
NUFTData = scipy.io.loadmat(BaseNUFTDataP + 'TrajForNUFT.mat')
Traj = NUFTData['Trajm2'][0:2, :]
# T=scipy.io.loadmat('/media/a/DATA/13May18/Me/meas_MID409_gBP_VD11_U19_7ADCs_FID17798/TrajForNUFT.mat')
# Traj=T['Trajm2'][0:2,:]
NMapCR = GT.GenerateNeighborsMap(Traj, kMax, aDataH, nccInData, ncc,
nNeighbors)
NMapCR = tf.constant(NMapCR)
# nNeighbors=myParams.myDict['nNeighbors']
# nccInData=myParams.myDict['nccInData']
# CurBartTraj=scipy.io.loadmat('/media/a/DATA/180628_AK/meas_MID244_gBP_VD11_U19_G35S155_4min_FID22439/Traj.mat')
# CurBartTraj=CurBartTraj['BARTTrajMS'][0:2,2:]
# osN=aDataH
# nNeighbors=nNeighbors
# NMap=np.zeros([osN,osN,nNeighbors],dtype='int32')
# C=GT.linspaceWithHalfStep(-kMax,kMax,osN)
# nChToUseInNN=nccInData
# ncc=nccInData
# nTrajAct=CurBartTraj.shape[1]
# for i in np.arange(0,osN):
# for j in np.arange(0,osN):
# CurLoc=np.vstack([C[i], C[j]])
# D=CurBartTraj-CurLoc
# R=np.linalg.norm(D,ord=2,axis=0)/np.sqrt(2)
# Idx=np.argsort(R)
# NMap[i,j,:]=Idx[0:nNeighbors]
# a=np.reshape(np.arange(0,nChToUseInNN)*nTrajAct,(1,1,1,nChToUseInNN))
# NMapC=np.reshape(NMap,(NMap.shape[0],NMap.shape[1],NMap.shape[2],1))+a
# NMapC=np.transpose(NMapC,(0,1,2,3))
# NMapCX=np.reshape(NMapC,(osN,osN,nNeighbors*nChToUseInNN))
# NMapCR=np.concatenate((NMapCX,NMapCX+nTrajAct*ncc),axis=2)
# model.print_shape()
# model.add_Reshape([16*133068])
model.add_Permute([1, 0, 2, 3]) # now we're 133068,16,1,1
# model.print_shape()
feature = model.get_output()
feature = tf.gather(feature, NMapCR, validate_indices=None,
name=None) # After 131,131,192,16
# feature = tf.reshape(feature, [aDataH, aDataW, achannelsIn])
model.add_PutInOutput(feature)
model.add_Permute([3, 0, 1, 2, 4, 5]) # After 16,131,131,192,1,1
# model.print_shape()
model.add_Reshape([batch_size, aDataH, aDataW,
achannelsIn]) # After 16,131,131,192
model.add_Split4thDim(
2) # Now we're kH,kW, Neighbors(12)*Channels(8),2
# model.add_PixelwiseMultC(nTS, stddev_factor=1.0) # After we're batch_size,kH,kW,nTS
InitForRC = []
if myParams.myDict['InitForRFN'] != 'None':
InitForRM = scipy.io.loadmat(myParams.myDict['InitForRFN'])
InitForRR = InitForRM['gene_GEN_L007_PixelwiseMultC_weightR_0']
InitForRI = InitForRM['gene_GEN_L007_PixelwiseMultC_weightI_0']
InitForRC = InitForRR + 1j * InitForRI
model.add_PixelwiseMultC(nTS,
stddev_factor=1.0,
NamePrefix='',
Trainable=True,
InitC=InitForRC)
MM = GT.gDFT_matrix(np.linspace(-kMax, kMax, aDataH), H)
MM = np.transpose(MM, axes=[1, 0])
if UseSharedWightesInRelaxedFT:
model.add_Mult2DMCyCSharedOverFeat(W,
1,
add_bias=addBias,
Trainable=RelaxedFT,
InitC=MM)
model.add_Mult2DMCxCSharedOverFeat(H,
1,
add_bias=addBias,
Trainable=RelaxedFT,
InitC=MM)
else:
model.add_Mult2DMCyC(W, 1, add_bias=addBias)
model.add_Mult2DMCxC(H, 1, add_bias=addBias)
# now supposedly batch_size,H,W,nTS
# model.add_PixelwiseMultC(1, stddev_factor=1.0) # This collecting the different TS to the final image.
InitForLC = []
if myParams.myDict['InitForLFN'] != 'None':
InitForLM = scipy.io.loadmat(myParams.myDict['InitForLFN'])
InitForLR = InitForLM['gene_GEN_L010_PixelwiseMultC_weightR_0']
InitForLI = InitForLM['gene_GEN_L010_PixelwiseMultC_weightI_0']
InitForLC = InitForLR + 1j * InitForLI
model.add_PixelwiseMultC(
1,
stddev_factor=1.0,
NamePrefix='',
Trainable=True,
InitC=InitForLC
) # This collecting the different TS to the final image.
model.remove_5thDim()
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == 'RegridTry3C2_TS_MB':
print("RegridTry3C2_TS_MB mode")
aDataH = myParams.myDict['aDataH']
kMax = myParams.myDict['kMax']
aDataW = myParams.myDict['aDataW']
# achannelsIn=myParams.myDict['achannelsIn']
nTS = myParams.myDict['nTimeSegments']
UseSharedWightesInRelaxedFT = myParams.myDict[
'UseSharedWightesInRelaxedFT'] > 0
RelaxedFT = myParams.myDict['RelaxedFT'] > 0
addBias = myParams.myDict['CmplxBias'] > 0
nccInData = myParams.myDict['nccInData']
ncc = myParams.myDict['nccToUse']
nNeighbors = myParams.myDict['nNeighbors']
achannelsIn = ncc * nNeighbors * 2
BaseNUFTDataP = myParams.myDict['BaseNUFTDataP']
NUFTData = scipy.io.loadmat(BaseNUFTDataP + 'TrajForNUFT.mat')
Traj = NUFTData['Trajm2'][0:2, :]
NMapCR = GT.GenerateNeighborsMap(Traj, kMax, aDataH, nccInData, ncc,
nNeighbors)
NMapCR = tf.constant(NMapCR)
model.add_Permute([1, 0, 2, 3]) # now we're 133068,16,1,1
# model.print_shape()
feature = model.get_output()
feature = tf.gather(feature, NMapCR, validate_indices=None,
name=None) # After 131,131,192,16
# feature = tf.reshape(feature, [aDataH, aDataW, achannelsIn])
model.add_PutInOutput(feature)
model.add_Permute([3, 0, 1, 2, 4, 5]) # After 16,131,131,192,1,1
# model.print_shape()
model.add_Reshape([batch_size, aDataH, aDataW,
achannelsIn]) # After 16,131,131,192
if addBias:
print("with bias")
else:
print("without bias")
model.add_Split4thDim(
2) # Now we're kH,kW, Neighbors(12)*Channels(8),2
# model.add_PixelwiseMultC(nTS, stddev_factor=1.0) # After we're batch_size,kH,kW,nTS
InitForRC = []
if myParams.myDict['InitForRFN'] != 'None':
InitForRM = scipy.io.loadmat(myParams.myDict['InitForRFN'])
InitForRR = InitForRM['gene_GEN_L007_PixelwiseMultC_weightR_0']
InitForRI = InitForRM['gene_GEN_L007_PixelwiseMultC_weightI_0']
InitForRC = InitForRR + 1j * InitForRI
model.add_PixelwiseMultC(nTS,
stddev_factor=1.0,
NamePrefix='',
Trainable=True,
InitC=InitForRC)
MM = GT.gDFT_matrix(np.linspace(-kMax, kMax, aDataH), H)
MM = np.transpose(MM, axes=[1, 0])
if UseSharedWightesInRelaxedFT:
model.add_Mult2DMCyCSharedOverFeat(W,
1,
add_bias=addBias,
Trainable=RelaxedFT,
InitC=MM)
model.add_Mult2DMCxCSharedOverFeat(H,
1,
add_bias=addBias,
Trainable=RelaxedFT,
InitC=MM)
else:
model.add_Mult2DMCyC(W, 1, add_bias=addBias)
model.add_Mult2DMCxC(H, 1, add_bias=addBias)
# now supposedly batch_size,H,W,nTS
# ggg: 2 here is MB
# model.add_PixelwiseMultC(2, stddev_factor=1.0) # This collecting the different TS to the final image.
# model.print_shape('BeforeL')
InitForLC = []
if myParams.myDict['InitForLFN'] != 'None':
InitForLM = scipy.io.loadmat(myParams.myDict['InitForLFN'])
InitForLR = InitForLM['gene_GEN_L010_PixelwiseMultC_weightR_0']
InitForLI = InitForLM['gene_GEN_L010_PixelwiseMultC_weightI_0']
InitForLC = InitForLR + 1j * InitForLI
model.add_PixelwiseMultC(
2,
stddev_factor=1.0,
NamePrefix='',
Trainable=True,
InitC=InitForLC
) # This collects the different TS to the final image.
# model.print_shape('AfterL')
model.add_Permute34()
model.add_Combine34(True)
# model.print_shape('After Combine34')
# model.remove_5thDim()
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == 'RegridTry1C2_TS':
print("RegridTry1C2_TS mode")
aDataH = myParams.myDict['aDataH']
kMax = myParams.myDict['kMax']
nTS = myParams.myDict['nTimeSegments']
UseSharedWightesInRelaxedFT = myParams.myDict[
'UseSharedWightesInRelaxedFT'] > 0
RelaxedFT = myParams.myDict['RelaxedFT'] > 0
addBias = myParams.myDict['CmplxBias'] > 0
model.add_Split4thDim(2)
# model.add_PixelwiseMultC(nTS, stddev_factor=1.0) # After we're batch_size,kH,kW,nTS
InitForRC = []
print("InitForRC...")
print(myParams.myDict['InitForRFN'])
if myParams.myDict['InitForRFN'] != 'None':
print("InitForRC From file")
InitForRM = scipy.io.loadmat(myParams.myDict['InitForRFN'])
InitForRR = InitForRM['gene_GEN_L007_PixelwiseMultC_weightR_0']
InitForRI = InitForRM['gene_GEN_L007_PixelwiseMultC_weightI_0']
InitForRC = InitForRR + 1j * InitForRI
print("InitForRC...")
model.add_PixelwiseMultC(nTS,
stddev_factor=1.0,
NamePrefix='',
Trainable=True,
InitC=InitForRC)
MM = GT.gDFT_matrix(np.linspace(-kMax, kMax, aDataH), H)
MM = np.transpose(MM, axes=[1, 0])
if UseSharedWightesInRelaxedFT:
model.add_Mult2DMCyCSharedOverFeat(W,
1,
add_bias=addBias,
Trainable=RelaxedFT,
InitC=MM)
model.add_Mult2DMCxCSharedOverFeat(H,
1,
add_bias=addBias,
Trainable=RelaxedFT,
InitC=MM)
else:
model.add_Mult2DMCyC(W, 1, add_bias=addBias)
model.add_Mult2DMCxC(H, 1, add_bias=addBias)
# now supposedly batch_size,H,W,nTS
# model.add_PixelwiseMultC(1, stddev_factor=1.0) # This collecting the different TS to the final image.
# add_PixelwiseMultC(self, numOutChannels, stddev_factor=1.0,NamePrefix='',Trainable=True,InitC=[]):
InitForLC = []
if myParams.myDict['InitForLFN'] != 'None':
InitForLM = scipy.io.loadmat(myParams.myDict['InitForLFN'])
InitForLR = InitForLM['gene_GEN_L010_PixelwiseMultC_weightR_0']
InitForLI = InitForLM['gene_GEN_L010_PixelwiseMultC_weightI_0']
InitForLC = InitForLR + 1j * InitForLI
model.add_PixelwiseMultC(
1,
stddev_factor=1.0,
NamePrefix='',
Trainable=True,
InitC=InitForLC
) # This collecting the different TS to the final image.
model.remove_5thDim()
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
# if myParams.myDict['NetMode'] == 'RegridTry1C2_TS2': # Shared features in relaxed FT
# print("RegridTry1C2_TS mode")
# addBias=myParams.myDict['CmplxBias']>0
# if addBias:
# print("with bias")
# else:
# print("without bias")
# nTS=7
# model.add_Split4thDim(2)
# model.add_PixelwiseMultC(nTS, stddev_factor=1.0)
# model.add_Mult2DMCyCSharedOverFeat(W,1,add_bias=addBias)
# model.add_Mult2DMCxCSharedOverFeat(H,1,add_bias=addBias)
# model.add_PixelwiseMultC(1, stddev_factor=1.0)
# model.remove_5thDim()
# new_vars = tf.global_variables()
# gene_vars = list(set(new_vars) - set(old_vars))
# return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == 'SMASHTry1':
print("SMASHTry1 mode")
addBias = myParams.myDict['CmplxBias'] > 0
model.add_PixelwiseMultC(2, stddev_factor=1.0)
model.add_Combine34()
model.add_Mult2DMCyC(W, 1, add_bias=addBias)
model.add_Mult2DMCxC(H, 1, add_bias=addBias)
model.remove_5thDim()
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == 'SMASHTry1_CC':
print("SMASHTry1_CC mode")
addBias = myParams.myDict['CmplxBias'] > 0
# we're [Batch, kH,kW,AllChannels*Neighbors*RI]
model.add_Split4thDim(2) # Now [Batch, kH,kW,AllChannels*Neighbors,RI]
model.add_Mult2DMCxCSharedOverFeat(
DataH, 1) # Now [Batch, H,kW,AllChannels*Neighbors,RI]
model.add_Split4thDim(6) # Now [Batch, H,kW,AllChannels,Neighbors,RI]
ncc = 4
model.add_einsumC('abcde,dx->abcxe', [8, ncc])
model.add_Combine45(
squeeze=True) # Now [Batch, H,kW,CompressedChannels*Neighbors,RI]
model.add_Mult2DMCxCSharedOverFeat(
DataH, 1) # Now [Batch, kH,kW,CompressedChannels*Neighbors,RI]
model.add_PixelwiseMultC(2, stddev_factor=1.0)
model.add_Combine34()
model.add_Mult2DMCyC(W, 1, add_bias=addBias)
model.add_Mult2DMCxC(H, 1, add_bias=addBias)
model.remove_5thDim()
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == 'SMASHTry1_GCC':
print("SMASHTry1_GCC mode")
addBias = myParams.myDict['CmplxBias'] > 0
# we're [Batch, kH,kW,AllChannels*Neighbors*RI]
model.add_Split4thDim(2) # Now [Batch, kH,kW,AllChannels*Neighbors,RI]
model.add_Mult2DMCxCSharedOverFeat(
DataH, 1) # Now [Batch, H,kW,AllChannels*Neighbors,RI]
model.add_Split4thDim(6) # Now [Batch, H,kW,AllChannels,Neighbors,RI]
ncc = 4
model.add_einsumC('abcde,bdx->abcxe', [DataH, 8, ncc])
model.add_Combine45(
squeeze=True) # Now [Batch, H,kW,CompressedChannels*Neighbors,RI]
model.add_Mult2DMCxCSharedOverFeat(
DataH, 1) # Now [Batch, kH,kW,CompressedChannels*Neighbors,RI]
model.add_PixelwiseMultC(2, stddev_factor=1.0)
model.add_Combine34()
model.add_Mult2DMCyC(W, 1, add_bias=addBias)
model.add_Mult2DMCxC(H, 1, add_bias=addBias)
model.remove_5thDim()
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == 'SMASHTry1_GCCF':
print("SMASHTry1_GCCF mode")
addBias = myParams.myDict['CmplxBias'] > 0
# we're [Batch, kH,kW,AllChannels*Neighbors*RI]
model.add_Split4thDim(2) # Now [Batch, H,kW,AllChannels*Neighbors,RI]
model.add_Split4thDim(6) # Now [Batch, H,kW,AllChannels,Neighbors,RI]
ncc = 4
model.add_einsumC('abcde,bdx->abcxe', [DataH, 8, ncc])
model.add_Combine45(
squeeze=True) # Now [Batch, H,kW,CompressedChannels*Neighbors,RI]
DFTM = DFT_matrix(DataH)
model.add_Mult2DMCxCSharedOverFeat(
DataH, 1, add_bias=addBias, Trainable=False,
InitC=DFTM) # Now [Batch, kH,kW,CompressedChannels*Neighbors,RI]
model.add_PixelwiseMultC(2, stddev_factor=1.0)
model.add_Combine34()
model.add_Mult2DMCyC(W, 1, add_bias=addBias)
model.add_Mult2DMCxC(H, 1, add_bias=addBias)
model.remove_5thDim()
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == 'Conv_3Layers':
print("Conv_3Layers mode")
# model.print_shape()
model.add_conv2d(64, mapsize=mapsize, stride=1, stddev_factor=2.)
model.add_elu()
model.add_conv2dWithName(32,
name="ggg",
mapsize=1,
stride=1,
stddev_factor=2.)
model.add_elu()
model.add_conv2d(channelsOut, mapsize=5, stride=1, stddev_factor=2.)
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == 'Unet_v1_ForB0':
b = np.array([[64, 0, 0, 128, 128, 0, 0, 64],
[128, 0, 0, 256, 256, 0, 0, 128],
[512, 0, 0, 0, 0, 0, 0, 512]])
model.add_UnetKsteps(b, mapsize=mapsize, stride=2, stddev_factor=1e-3)
# OutChannels=labels.shape[3]
model.add_conv2d(channelsOut,
mapsize=mapsize,
stride=1,
stddev_factor=2.)
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == 'Conv_v1_ForB0':
print("Conv_v1_ForB0 mode")
# model.print_shape()
model.add_conv2d(64, mapsize=mapsize, stride=1, stddev_factor=2.)
model.add_elu()
model.add_conv2dWithName(128,
name="ggg",
mapsize=mapsize,
stride=1,
stddev_factor=2.)
model.add_elu()
model.add_conv2d(128, mapsize=mapsize, stride=1, stddev_factor=2.)
model.add_elu()
model.add_conv2d(channelsOut,
mapsize=mapsize,
stride=1,
stddev_factor=2.)
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
if myParams.myDict['NetMode'] == 'Conv_v1':
print("Conv_v1 mode")
# model.print_shape()
model.add_conv2d(64, mapsize=mapsize, stride=1, stddev_factor=2.)
model.add_elu()
model.add_conv2dWithName(128,
name="ggg",
mapsize=mapsize,
stride=1,
stddev_factor=2.)
model.add_elu()
model.add_conv2d(128, mapsize=mapsize, stride=1, stddev_factor=2.)
model.add_elu()
model.add_conv2d(1, mapsize=mapsize, stride=1, stddev_factor=2.)
# SAE
SAE = myParams.myDict['NetMode'] == 'SAE'
if SAE:
model.add_conv2d(64, mapsize=mapsize, stride=1, stddev_factor=2.)
model.add_elu()
model.add_conv2dWithName(128,
name="AE",
mapsize=mapsize,
stride=1,
stddev_factor=2.)
model.add_conv2d(64, mapsize=mapsize, stride=1, stddev_factor=2.)
model.add_elu()
model.add_conv2d(channels, mapsize=7, stride=1, stddev_factor=2.)
model.add_sigmoid()
# model.add_tanh()
# kKick:
kKick = myParams.myDict['NetMode'] == 'kKick'
if kKick:
model.add_conv2d(64, mapsize=1, stride=1, stddev_factor=2.)
model.add_elu()
b = np.array([[64, 0, 0, 128, 128, 0, 0, 64],
[128, 0, 0, 256, 256, 0, 0, 128],
[512, 0, 0, 0, 0, 0, 0, 512]])
model.add_UnetKsteps(b, mapsize=mapsize, stride=2, stddev_factor=1e-3)
model.add_conv2dWithName(50,
name="AE",
mapsize=3,
stride=1,
stddev_factor=2.)
model.add_elu()
model.add_conv2d(channels, mapsize=1, stride=1, stddev_factor=2.)
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
return model.get_output(), gene_vars
# AUTOMAP
# AUTOMAP_units = [64, 64, channels]
# AUTOMAP_mapsize = [5, 5, 7]
# ggg option 1: FC
# model.add_flatten() # FC1
# model.add_dense(num_units=H*W*2)
# model.add_reshapeTo4D(H,W)
TSRECON = myParams.myDict['NetMode'] == 'TSRECON'
if TSRECON:
# ggg option 2: FC per channel, and then dot multiplication per pixel, then conv
ChannelsPerCoil = myParams.myDict['NumFeatPerChannel']
NumTotalFeat = myParams.myDict['NumTotalFeat']
model.add_Mult2DMC(H * W, ChannelsPerCoil)
model.add_reshapeTo4D(H, W)
model.add_PixelwiseMult(NumTotalFeat, stddev_factor=1.0)
model.add_elu()
#model.add_denseFromM('piMDR')
#model.add_reshapeTo4D(FLAGS.LabelsH,FLAGS.LabelsW)
# #model.add_tanh() # FC2
#model.add_Unet1Step(128, mapsize=5, stride=2, num_layers=2, stddev_factor=1e-3)
#model.add_conv2d(channels, mapsize=5, stride=1, stddev_factor=2.)
b = np.array([[64, 0, 0, 128, 128, 0, 0, 64],
[128, 0, 0, 256, 256, 0, 0, 128],
[512, 0, 0, 0, 0, 0, 0, 512]])
#b=np.array([[64,0,0,128,128,0,0,64],[128,0,0,256,256,0,0,128]])
#b=np.array([[64,0,0,0,0,0,0,64]])
model.add_UnetKsteps(b, mapsize=mapsize, stride=2, stddev_factor=1e-3)
# model.add_conv2d(channels, mapsize=1, stride=1, stddev_factor=2.)
# ggg: Autoencode
model.add_conv2dWithName(50,
name="AE",
mapsize=3,
stride=1,
stddev_factor=2.)
model.add_elu()
# ggg: Finish
model.add_conv2d(channels, mapsize=1, stride=1, stddev_factor=2.)
# #model.add_flatten()
# #model.add_dense(num_units=H*W*1)
# model.add_reshapeTo4D(FLAGS.LabelsH,FLAGS.LabelsW)
# #model.add_batch_norm()
# #model.add_tanh() # TC3
# # model.add_conv2d(AUTOMAP_units[0], mapsize=AUTOMAP_mapsize[0], stride=1, stddev_factor=2.)
# # model.add_batch_norm()
# #model.add_relu()
# #model.add_conv2d(AUTOMAP_units[1], mapsize=AUTOMAP_mapsize[1], stride=1, stddev_factor=2.)
# # model.add_batch_norm()
# #model.add_relu()
# #model.add_conv2d(AUTOMAP_units[2], mapsize=AUTOMAP_mapsize[2], stride=1, stddev_factor=2.)
# # model.add_conv2d(AUTOMAP_units[2], mapsize=1, stride=1, stddev_factor=2.)
# # model.add_relu()
#model.add_constMatMul()
#for ru in range(len(res_units)-1):
# nunits = res_units[ru]
# for j in range(2):
# model.add_residual_block(nunits, mapsize=mapsize)
# Spatial upscale (see http://distill.pub/2016/deconv-checkerboard/)
# and transposed convolution
# model.add_upscale()
# model.add_batch_norm()
# model.add_relu()
# model.add_conv2d_transpose(nunits, mapsize=mapsize, stride=1, stddev_factor=1.)
# model.add_flatten()
# model.add_dense(num_units=H*W*4)
# model.add_reshapeTo4D(FLAGS.LabelsH,FLAGS.LabelsW)
# #model.add_Mult2D()
# #model.add_Mult3DComplexRI()
SrezOrigImagePartModel = False
if SrezOrigImagePartModel:
nunits = res_units[0]
for j in range(2):
model.add_residual_block(nunits, mapsize=mapsize)
#model.add_upscale()
model.add_batch_norm()
model.add_relu()
model.add_conv2d_transpose(nunits,
mapsize=mapsize,
stride=1,
stddev_factor=1.)
nunits = res_units[1]
for j in range(2):
model.add_residual_block(nunits, mapsize=mapsize)
#model.add_upscale()
model.add_batch_norm()
model.add_relu()
model.add_conv2d_transpose(nunits,
mapsize=mapsize,
stride=1,
stddev_factor=1.)
# Finalization a la "all convolutional net"
nunits = res_units[-1]
model.add_conv2d(nunits, mapsize=mapsize, stride=1, stddev_factor=2.)
# Worse: model.add_batch_norm()
model.add_relu()
model.add_conv2d(nunits, mapsize=1, stride=1, stddev_factor=2.)
# Worse: model.add_batch_norm()
model.add_relu()
# Last layer is sigmoid with no batch normalization
model.add_conv2d(channels, mapsize=1, stride=1, stddev_factor=1.)
model.add_sigmoid()
new_vars = tf.global_variables()
gene_vars = list(set(new_vars) - set(old_vars))
# ggg = tf.identity(model.get_output(), name="ggg")
return model.get_output(), gene_vars