def LoadData(self, updatefunction=None):
'Called when first starting the program, or when preference change sets dirtyDicts'
config = self.config
if config['charset']:
charset = config['charset']
else:
charset = 'simplified'
self.dicts = []
for dictname in config['dictionaries']:
self.addMessage("Loading dictionary %s ..." % dictname)
dictFile = os.path.join(config.appDir, 'dict', dictname)
dict = segmenter.Dictionary(dictFile,
format='cedict',
verbose=True,
updatefunction=updatefunction)
if dict.messages != None:
for elem in dict.messages:
self.addMessage(elem)
# add a blank line
self.addMessage("Loaded dictionary %s, %d words" %
(dictname, dict.getWordCount()))
self.dicts.append(dict)
self.seg = segmenter.Segmenter(charset, self.dicts, self.stats)
self.addMessage("")
def main(tds, vds):
# setup experiment logging to comet.ml
if expConfig.LOG_COMETML:
hyper_params = {
"experimentName": expConfig.EXPERIMENT_NAME,
"epochs": expConfig.EPOCHS,
"batchSize": expConfig.BATCH_SIZE,
"channels": expConfig.CHANNELS,
"virualBatchsize": expConfig.VIRTUAL_BATCHSIZE
}
expConfig.experiment.log_parameters(hyper_params)
expConfig.experiment.add_tags(
[expConfig.EXPERIMENT_NAME, "ID{}".format(expConfig.id)])
if hasattr(expConfig, "EXPERIMENT_TAGS"):
expConfig.experiment.add_tags(expConfig.EXPERIMENT_TAGS)
print(bcolors.OKGREEN + "Logging to comet.ml" + bcolors.ENDC)
else:
print(bcolors.WARNING + "Not logging to comet.ml" + bcolors.ENDC)
# log parameter count
if expConfig.LOG_PARAMCOUNT:
paramCount = sum(p.numel() for p in expConfig.net.parameters()
if p.requires_grad)
print("Parameters: {:,}".format(paramCount).replace(",", "'"))
#load data
randomCrop = expConfig.RANDOM_CROP if hasattr(expConfig,
"RANDOM_CROP") else None
trainloader = torch.utils.data.DataLoader(
tds,
batch_size=expConfig.BATCH_SIZE,
shuffle=True,
pin_memory=True,
num_workers=expConfig.DATASET_WORKERS)
valloader = torch.utils.data.DataLoader(
vds,
batch_size=1,
shuffle=False,
pin_memory=True,
num_workers=expConfig.DATASET_WORKERS)
challengeValloader = torch.utils.data.DataLoader(
vds,
batch_size=1,
shuffle=False,
pin_memory=True,
num_workers=expConfig.DATASET_WORKERS)
seg = segmenter.Segmenter(expConfig, trainloader, valloader,
challengeValloader)
if hasattr(expConfig, "VALIDATE_ALL") and expConfig.VALIDATE_ALL:
seg.validateAllCheckpoints()
elif hasattr(expConfig, "PREDICT") and expConfig.PREDICT:
seg.makePredictions()
else:
seg.train()
def main():
'''Get data as a string from read_in() and write out the segments'''
line = read_in()
s = seg.Segmenter()
segments = get_segments(line, s)
for s in flatten(segments):
print("%s%s" % (s[0].upper(), s[1:]) + '.')
def main(args):
expConfig=exps[args.exp]
# setup experiment logging to comet.ml
if expConfig.LOG_COMETML:
hyper_params = {"experimentName": expConfig.EXPERIMENT_NAME,
"epochs": expConfig.EPOCHS,
"batchSize": expConfig.BATCH_SIZE,
"channels": expConfig.CHANNELS,
"virualBatchsize": expConfig.VIRTUAL_BATCHSIZE}
expConfig.experiment.log_(hyper_params)
expConfig.experiment.add_tags([expConfig.EXPERIMENT_NAME, "ID{}".format(expConfig.id)])
if hasattr(expConfig, "EXPERIMENT_TAGS"): expConfig.experiment.add_tags(expConfig.EXPERIMENT_TAGS)
print(bcolors.OKGREEN + "Logging to comet.ml" + bcolors.ENDC)
else:
print(bcolors.WARNING + "Not logging to comet.ml" + bcolors.ENDC)
# log parameter count
if expConfig.LOG_PARAMCOUNT:
paramCount = sum(p.numel() for p in expConfig.net.parameters() if p.requires_grad)
print("Parameters: {:,}".format(paramCount).replace(",", "'"))
#load data
randomCrop = None
trainset = bratsDataset.BratsDataset(systemsetup.BRATS_PATH, expConfig, mode="train", randomCrop=randomCrop)
trainloader = torch.utils.data.DataLoader(trainset, batch_size=expConfig.BATCH_SIZE, shuffle=True, pin_memory=False)
valset = bratsDataset.BratsDataset(systemsetup.BRATS_PATH, expConfig, mode="validation")
valloader = torch.utils.data.DataLoader(valset, batch_size=1, shuffle=False, pin_memory=False)
challengeValset = bratsDataset.BratsDataset(systemsetup.BRATS_VAL_PATH, expConfig,hasMasks=False, mode="validation", returnOffsets=True)
challengeValloader = torch.utils.data.DataLoader(challengeValset, batch_size=1, shuffle=False, pin_memory=True)
seg = segmenter.Segmenter(expConfig, trainloader, valloader, challengeValloader)#,trainvalloader)
if hasattr(expConfig, "VALIDATE_ALL") and expConfig.VALIDATE_ALL:
seg.validateAllCheckpoints()
elif hasattr(expConfig, "PREDICT") and expConfig.PREDICT:
seg.makePredictions()
else:
seg.train()
gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
processedimage = cv2.adaptiveThreshold(gray, 255,
cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
cv2.THRESH_BINARY, 11, 2)
processedimage = cv2.medianBlur(processedimage, 3)
return processedimage
def deskew_image(im, pim):
cornerdetector = cd.CornerDetector(pim)
corners = cornerdetector.detect_corners()
for i in range(0, len(corners)):
corner = corners[i]
nextcorner = corners[(i + 1) % len(corners)]
cv2.circle(im, (corner[0], corner[1]), 25, (210, 210, 210), -1)
cv2.circle(im, (corner[0], corner[1]), 27, (130, 210, 210), 4)
cv2.line(im, tuple(corner), tuple(nextcorner), (130, 210, 210), 5)
plt.imshow(im, cmap='gray')
plt.show()
deskewer = deskew.Deskewer(im, corners, 1.414)
transformedimage = deskewer.deskew()
return transformedimage
img = cv2.cvtColor(cv2.imread('form5.jpg'), cv2.COLOR_BGR2RGB)
im = preprocess_image(img)
img = deskew_image(img, im)
pim = preprocess_image(img)
segm = seg.Segmenter(img, pim)
segm.display_segments()
def main():
# setup experiment logging to comet.ml
if expConfig.LOG_COMETML:
hyper_params = {
"experimentName": expConfig.EXPERIMENT_NAME,
"epochs": expConfig.EPOCHS,
"batchSize": expConfig.BATCH_SIZE,
"channels": expConfig.CHANNELS,
"virualBatchsize": expConfig.VIRTUAL_BATCHSIZE
}
expConfig.experiment.log_parameters(hyper_params)
expConfig.experiment.add_tags(
[expConfig.EXPERIMENT_NAME, "ID{}".format(expConfig.id)])
if hasattr(expConfig, "EXPERIMENT_TAGS"):
expConfig.experiment.add_tags(expConfig.EXPERIMENT_TAGS)
print(bcolors.OKGREEN + "Logging to comet.ml" + bcolors.ENDC)
else:
print(bcolors.WARNING + "Not logging to comet.ml" + bcolors.ENDC)
# log parameter count
if expConfig.LOG_PARAMCOUNT:
paramCount = sum(p.numel() for p in expConfig.net.parameters()
if p.requires_grad)
print("Parameters: {:,}".format(paramCount).replace(",", "'"))
# load data
if expConfig.AVERAGE_DATA:
average_data, _, _ = utils.load_nii(
"/home/liujing/data/MICCAI_BraTS/2019/training/MixupData.nii.gz")
else:
average_data = None
randomCrop = expConfig.RANDOM_CROP if hasattr(expConfig,
"RANDOM_CROP") else None
trainset = bratsDataset.BratsDataset(systemsetup.BRATS_PATH,
expConfig,
mode="train",
randomCrop=randomCrop,
average_data=average_data)
trainloader = torch.utils.data.DataLoader(
trainset,
batch_size=expConfig.BATCH_SIZE,
shuffle=True,
pin_memory=True,
num_workers=expConfig.DATASET_WORKERS)
valset = bratsDataset.BratsDataset(systemsetup.BRATS_PATH,
expConfig,
mode="validation",
average_data=average_data)
valloader = torch.utils.data.DataLoader(
valset,
batch_size=1,
shuffle=False,
pin_memory=True,
num_workers=expConfig.DATASET_WORKERS)
challengeValset = bratsDataset.BratsDataset(systemsetup.BRATS_VAL_PATH,
expConfig,
mode="validation",
hasMasks=False,
returnOffsets=True,
average_data=average_data)
challengeValloader = torch.utils.data.DataLoader(
challengeValset,
batch_size=1,
shuffle=False,
pin_memory=True,
num_workers=expConfig.DATASET_WORKERS)
seg = segmenter.Segmenter(expConfig, trainloader, valloader,
challengeValloader, trainset)
if hasattr(expConfig, "VALIDATE_ALL") and expConfig.VALIDATE_ALL:
seg.validateAllCheckpoints()
elif hasattr(expConfig, "PREDICT") and expConfig.PREDICT:
seg.makePredictions()
else:
seg.train(is_mixup=False)