def continue_em_training(epochNumber):
initialEpoch = 103
netRoot = "/home/oole/lymphoma_net/"
runName = BASENAME + "_em_" + EMRUNSTAMP + "_cont/"
modelName = "lymph_model"
if not os.path.exists(netRoot):
os.makedirs(netRoot)
else:
print("Net root folder already extists.")
if not os.path.exists(netRoot + runName):
os.makedirs(netRoot + runName)
else:
print("Run folder already extists.")
old_em_savepath = netRoot + runName + BASENAME + "_em"
em_train_savepath = netRoot + runName + BASENAME + "_em"
logfile_path = netRoot + runName + BASENAME + "_net_log_em.csv"
logreg_savepath = netRoot + runName + BASENAME + "_logreg"
# load data
# split into train val
dataPath = "/home/oole/data_lymphoma/"
trainSlideData = ldata.collect_data(dataPath)
valSlideData = data_tf.SlideData(
trainSlideData.getSlideList(),
None,
np.asarray(trainSlideData.getSlideList()).size,
trainSlideData.getSlideLabelList(),
trainSlideData.getLabelFunc(),
False,
labelencoder=trainSlideData.getLabelEncoder(),
parseFunctionAugment=trainSlideData.getparseFunctionNormal(),
parseFunction=trainSlideData.getparseFunctionNormal())
train_em.emtrain(trainSlideData,
valSlideData,
old_em_savepath,
em_train_savepath,
BATCH_SIZE,
initial_epochnum=initialEpoch,
model_name=modelName,
spatial_smoothing=SPATIALSMOOTHING,
do_augment=True,
num_epochs=epochNumber,
dropout_ratio=DROPOUT_RATIO,
learning_rate=LEARNING_RATE,
sanity_check=False,
logfile_path=logfile_path,
logreg_savepath=logreg_savepath,
runName=runName,
netAcc=None,
buildNet=lnet.getLymphNet,
valIsTestData=True,
discriminativePatchFinderTrain=TRAIN_DISC_FINDER,
discriminativePatchFinderPredict=PRED_DISC_FINDER,
splitSeed=SPLIT_SEED)
print("Data collected.")
def continue_em_training():
initialEpoch = 103
netRoot = "/home/oole/lymphoma_net/"
runName = "lymphoma_em_180815_2_cont/"
modelName = "lymph_model"
if not os.path.exists(netRoot):
os.makedirs(netRoot)
else:
print("Net root folder already extists.")
if not os.path.exists(netRoot + runName):
os.makedirs(netRoot + runName)
else:
print("Run folder already extists.")
old_em_savepath = netRoot + "lymphoma_em_180815_2/" + "lympf_em"
em_train_savepath = netRoot + runName + "lympf_em"
logfile_path = netRoot + runName + "lymph_net_log_em.csv"
logreg_savepath = netRoot + runName + "lymph_logreg"
# load data
# split into train val
dataPath = "/home/oole/data_lymphoma/"
trainSlideData = ldata.collect_data(dataPath)
valSlideData = data_tf.SlideData(
trainSlideData.getSlideList(),
None,
np.asarray(trainSlideData.getSlideList()).size,
trainSlideData.getSlideLabelList(),
trainSlideData.getLabelFunc(),
False,
labelencoder=trainSlideData.getLabelEncoder(),
parseFunctionAugment=trainSlideData.getparseFunctionNormal(),
parseFunction=trainSlideData.getparseFunctionNormal())
train_em.emtrain(trainSlideData,
valSlideData,
old_em_savepath,
em_train_savepath,
BATCH_SIZE,
initial_epochnum=initialEpoch,
model_name=modelName,
spatial_smoothing=SPATIALSMOOTHING,
do_augment=True,
num_epochs=100,
dropout_ratio=DROPOUT_RATIO,
learning_rate=LEARNING_RATE,
sanity_check=False,
logfile_path=logfile_path,
logreg_savepath=logreg_savepath,
runName=runName,
netAcc=None,
buildNet=lnet.getLymphNet)
print("Data collected.")
def train_augment():
# simple_train_savepath = "/home/oole/tfnetsave/tfnet_full"
simple_train_savepath = "/home/oole/tfnetsave/tfnet_em_full"
em_train_savepath = "/home/oole/tfnetsave/tfnet_em_full"
initial_epoch = 0
train_datapath = "/home/oole/Data/training/patient_patches_jpg"
# train_datapath = '/home/oole/tf_test_data/validation'
val_datapath = "/home/oole/Data/validation/patient_patches_jpg"
logfile_path = "/home/oole/tfnetsave/tfnet_em_full_log.csv"
logreg_savepath = "/home/oole/tfnetsave/tfnet_em_full_logreg"
model_name = "model"
label_encoder = data_tf.labelencoder()
train_slidelist, train_slide_dimensions, old_disc_patches, _ = data_tf.collect_data(
train_datapath, batch_size)
val_slidelist, _, _, _ = data_tf.collect_data(val_datapath, batch_size)
train_patches = dataset.slidelist_to_patchlist(train_slidelist)
val_patches = dataset.slidelist_to_patchlist(val_slidelist)
np.random.shuffle(train_patches)
np.random.shuffle(val_patches)
# Initial training
#train_accuracy, val_accuracy = train.train_net(train_patches, val_patches, num_epochs=2, batch_size=batch_size,
# savepath=simple_train_savepath, do_augment=True, model_name=model_name)
#util.write_log_file(logfile_path, train_accuracy=train_accuracy, val_accuracy=val_accuracy)
# Test continue training
# train.train_net(train_patches, val_patches, num_epochs=2, batch_size=batch_size, savepath=simple_train_savepath,
# loadpath=simple_train_savepath, do_augment=False, model_name="model")
train_em.emtrain(train_datapath,
val_datapath,
simple_train_savepath,
em_train_savepath,
label_encoder,
batch_size,
initial_epochnum=initial_epoch,
model_name=model_name,
spatial_smoothing=False,
do_augment=True,
num_epochs=2,
dropout_ratio=0.5,
learning_rate=0.0005,
sanity_check=False,
logfile_path=logfile_path,
logreg_savepath=logreg_savepath)
def em_training(initialEpoch=2, epochNumber=198):
netRoot = "/home/oole/lymphoma_net_vgg/"
modelName = BASENAME + "_model"
runName = BASENAME + "_em_" + EMRUNSTAMP + "/"
if not os.path.exists(netRoot):
os.makedirs(netRoot)
else:
print("Net root folder already extists.")
if not os.path.exists(netRoot + runName):
os.makedirs(netRoot + runName)
else:
print("Run folder already extists.")
simple_train_loadpath = netRoot + BASENAME + "_simple_" + SIMPLERUNSTAMP + "/" + BASENAME + "_simple"
em_train_savepath = netRoot + runName + BASENAME + "_em"
logfile_path = netRoot + runName + BASENAME + "_net_log_em.csv"
logreg_savepath = netRoot + runName + BASENAME + "_logreg"
# load data
# split into train val
basePath = "/home/oole/data_lymphoma/"
trainDataPath = basePath + "train/"
testDataPath = basePath + "test/"
trainSlideData = ldata.collect_data(trainDataPath)
testSlideData = ldata.collect_data(testDataPath)
train_em.emtrain(trainSlideData,
testSlideData,
simple_train_loadpath,
em_train_savepath,
BATCH_SIZE,
initial_epochnum=initialEpoch,
model_name=modelName,
spatial_smoothing=SPATIALSMOOTHING,
do_augment=True,
num_epochs=epochNumber,
dropout_ratio=DROPOUT_RATIO,
learning_rate=LEARNING_RATE,
sanity_check=False,
logfile_path=logfile_path,
logreg_savepath=logreg_savepath,
runName=runName,
netAcc=None,
buildNet=lnet.getLymphNet,
valIsTestData=True,
discriminativePatchFinderTrain=TRAIN_DISC_FINDER,
discriminativePatchFinderPredict=PRED_DISC_FINDER,
splitSeed=SPLIT_SEED)
print("Data collected.")
def em_training():
initialEpoch = 2
trainepochs = 200
netRoot = "/home/oole/lymphoma_net/"
runName = "lymphoma_em_180910/"
modelName = "lymph_model"
if not os.path.exists(netRoot):
os.makedirs(netRoot)
else:
print("Net root folder already extists.")
if not os.path.exists(netRoot + runName):
os.makedirs(netRoot + runName)
else:
print("Run folder already extists.")
simple_train_loadpath = netRoot + "lymphoma_simple_180910/" + "lymph_simple"
em_train_savepath = netRoot + runName + "lympf_em"
logfile_path = netRoot + runName + "lymph_net_log_em.csv"
logreg_savepath = netRoot + runName + "lymph_logreg"
# load data
# split into train val
basePath = "/home/oole/data_lymphoma/"
trainDataPath = basePath + "train/"
testDataPath = basePath + "test/"
trainSlideData = ldata.collect_data(trainDataPath)
testSlideData = ldata.collect_data(testDataPath)
train_em.emtrain(trainSlideData,
testSlideData,
simple_train_loadpath,
em_train_savepath,
BATCH_SIZE,
initial_epochnum=initialEpoch,
model_name=modelName,
spatial_smoothing=SPATIALSMOOTHING,
do_augment=True,
num_epochs=trainepochs,
dropout_ratio=DROPOUT_RATIO,
learning_rate=LEARNING_RATE,
sanity_check=False,
logfile_path=logfile_path,
logreg_savepath=logreg_savepath,
runName=runName,
netAcc=None,
buildNet=lnet.getLymphNet,
valIsTestData=True)
print("Data collected.")
def em_training():
initialEpoch = 2
netRoot = "/home/oole/breasthistology_net/"
runName = "breasthistology_em_180815_2/"
modelName = "breasthistology_model"
if not os.path.exists(netRoot):
os.makedirs(netRoot)
else:
print("Net root folder already extists.")
if not os.path.exists(netRoot + runName):
os.makedirs(netRoot + runName)
else:
print("Run folder already extists.")
simple_train_loadpath = netRoot + "breasthistology_simple_180815/" + "breasthistology_simple"
em_train_savepath = netRoot + runName + "breasthistology_em"
logfile_path = netRoot + runName + "breasthistology_net_log_em.csv"
logreg_savepath = netRoot + runName + "breasthistology_logreg"
# load data
# split into train val
trainDataPath = "/home/oole/breasthistology/training/"
testDataPath = "/home/oole/breasthistology/testing/"
trainSlideData = bdata.collect_data(trainDataPath)
testSlideData = bdata.collect_data(testDataPath)
train_em.emtrain(trainSlideData,
testSlideData,
simple_train_loadpath,
em_train_savepath,
BATCH_SIZE,
initial_epochnum=initialEpoch,
model_name=modelName,
spatial_smoothing=SPATIALSMOOTHING,
do_augment=True,
num_epochs=100,
dropout_ratio=DROPOUT_RATIO,
learning_rate=LEARNING_RATE,
sanity_check=False,
logfile_path=logfile_path,
logreg_savepath=logreg_savepath,
runName=runName,
netAcc=None,
buildNet=bnet.getBreasthistoNet,
valIsTestData=True)
print("Data collected.")
def cross_val_training(foldToTrain, numberOfEpochs=30):
initialEpoch = 0
netRoot = "/home/oole/lymphoma_net_vgg/"
modelName = BASENAME + "_model"
if not os.path.exists(netRoot):
os.makedirs(netRoot)
else:
print("Net root folder already extists.")
# load data
# split into train val
basePath = "/home/oole/data_lymphoma_full/"
trainDataPath = basePath + "train/"
# testDataPath = basePath + "test/"
trainSlideData = ldata.collect_data(trainDataPath)
testSlideData = trainSlideData
sess = tf.Session()
netAcc = None
foldNum = 1
for trainSlideData, testSLideData in data_tf.KFoldSlideList(
trainSlideData, testSlideData, numberOfSplits=5, splitSeed=1337):
print("Fold number: " + str(foldNum))
if foldNum == foldToTrain:
# train slide data should consist of all available training data
# is split into test and training
runName = BASENAME + "_simple_" + SIMPLERUNSTAMP + "_k-" + str(
foldNum) + "-5" + "/"
if not os.path.exists(netRoot + runName):
os.makedirs(netRoot + runName)
else:
print("Run folder already extists.")
train_savepath = netRoot + runName + BASENAME + "_k-" + str(
foldNum) + "-5"
logreg_savepath = netRoot + runName + BASENAME + "_logreg" + "_k-" + str(
foldNum) + "-5"
logfile_path = netRoot + runName + BASENAME + "_k-" + str(
foldNum) + "-5" + "_net_log_em.csv"
if netAcc is not None:
netAcc.getSummmaryWriter(runName, sess.graph, forceNew=True)
_, _, netAcc = train.train_net(trainSlideData,
testSlideData,
num_epochs=2,
batch_size=BATCH_SIZE,
savepath=train_savepath,
do_augment=True,
model_name=modelName,
getlabel_train=ldata.getlabel,
runName=runName,
lr=LEARNING_RATE,
buildNet=lnet.getLymphNet,
valIsTestData=True,
splitSeed=SPLIT_SEED,
sess=sess,
netAcc=netAcc,
initialEpoch=initialEpoch,
verbose=1,
do_simple_validation=False)
print("Finished Simple Training")
train_em.emtrain(trainSlideData,
testSlideData,
None,
train_savepath,
BATCH_SIZE,
initial_epochnum=initialEpoch + 2,
model_name=modelName,
spatial_smoothing=SPATIALSMOOTHING,
do_augment=True,
num_epochs=initialEpoch + numberOfEpochs,
dropout_ratio=DROPOUT_RATIO,
learning_rate=LEARNING_RATE,
sanity_check=False,
logreg_savepath=logreg_savepath,
runName=runName,
netAcc=netAcc,
valIsTestData=True,
discriminativePatchFinderTrain=TRAIN_DISC_FINDER,
discriminativePatchFinderPredict=PRED_DISC_FINDER,
splitSeed=SPLIT_SEED,
sess=sess,
verbose=1,
do_simple_validation=False)
foldNum = foldNum + 1
