import config
from util import data_loader
from util import losses
from util import keras_util
from util.keras_util import KerasModelConfig
model_config = KerasModelConfig(
k_fold_file="1.txt",
model_path=os.path.abspath(__file__),
image_resolution=400,
data_type=[config.DATA_TYPE_ORIGINAL],
label_position=[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12],
label_color_augment=[0, 1, 3, 5, 6, 7, 9, 10, 11, 12],
train_batch_size=[16, 16, 16],
label_up_sampling=[10, 0, 0, 10, 0, 0, 10, 0, 0, 0, 0, 0, 10],
data_visualization=True,
initial_epoch=4,
downsampling=0.5,
val_batch_size=128,
predict_batch_size=128,
epoch=[1, 3, 5],
lr=[0.001, 0.0001, 0.00001],
freeze_layers=[-1, 0.6, 5],
tta_flip=True,
input_norm=False)
def get_model(freeze_layers=-1, lr=0.01, output_dim=1, weights="imagenet"):
base_model = keras.applications.ResNet50(
include_top=False,
weights=weights,
from keras.layers import Dense, BatchNormalization, Activation
import config
from util import data_loader
from util import keras_util
from util import losses
from util.keras_util import KerasModelConfig
model_config = KerasModelConfig(k_fold_file="1.txt",
model_path=os.path.abspath(__file__),
image_resolution=224,
data_type=[config.DATA_TYPE_ORIGINAL],
label_position=[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12],
train_batch_size=[32, 32],
data_visualization=True,
val_batch_size=256,
predict_batch_size=256,
epoch=[2, 10],
lr=[0.0001, 0.00001],
clr=True,
freeze_layers=[0.1, 0.1],
tta_flip=True,
input_norm=False)
def get_model(freeze_layers=-1, lr=0.01, output_dim=1, weights="imagenet"):
base_model = keras.applications.DenseNet169(include_top=False, weights=weights,
input_shape=model_config.image_shape, pooling="avg")
x = base_model.output
x = Dense(256, use_bias=False)(x)
import keras
from keras.layers import Dense, BatchNormalization, Activation
import config
from util import data_loader
from util import keras_util
from util.keras_util import KerasModelConfig
model_config = KerasModelConfig(
k_fold_file="1.txt",
model_path=os.path.abspath(__file__),
image_resolution=224,
data_type=[config.DATA_TYPE_ORIGINAL],
label_position=[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12],
label_up_sampling=[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 15, 0],
train_batch_size=[32, 32, 32],
val_batch_size=256,
predict_batch_size=256,
epoch=[1, 4, 10],
lr=[0.001, 0.0001, 0.00001],
data_visualization=True,
freeze_layers=[-1, 0.6, 5],
tta_flip=True,
tta_crop=True)
def get_model(freeze_layers=-1, lr=0.01, output_dim=1, weights="imagenet"):
base_model = keras.applications.DenseNet169(
include_top=False,
weights=weights,
input_shape=model_config.image_shape,
pooling="avg")
import keras
from keras.layers import Dense, BatchNormalization, Activation
import config
from util import data_loader
from util import keras_util
from util.keras_util import KerasModelConfig
model_config = KerasModelConfig(
k_fold_file="1.txt",
model_path=os.path.abspath(__file__),
image_resolution=224,
data_type=[config.DATA_TYPE_ORIGINAL],
label_position=[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12],
label_color_augment=[0, 1, 3, 5, 6, 7, 9, 10, 11, 12],
train_batch_size=[16, 16, 16],
initial_epoch=1,
val_batch_size=256,
predict_batch_size=256,
epoch=[1, 4, 10],
lr=[0.0005, 0.00005, 0.000005],
data_visualization=True,
freeze_layers=[-1, 0.6, 5])
def get_model(freeze_layers=-1, lr=0.01, output_dim=1, weights="imagenet"):
base_model = keras.applications.DenseNet201(
include_top=False,
weights=weights,
input_shape=model_config.image_shape,
pooling="avg")
import time
import keras
from keras.layers import Dense, BatchNormalization, Activation
import config
from util import data_loader
from util import keras_util
from util.keras_util import KerasModelConfig
model_config = KerasModelConfig(k_fold_file="1.txt",
model_path=os.path.abspath(__file__),
image_resolution=224,
data_type=[config.DATA_TYPE_ORIGINAL],
label_position=[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12],
train_batch_size=[16, 16, 16],
val_batch_size=256,
predict_batch_size=256,
epoch=[1, 4, 12],
lr=[0.001, 0.0001, 0.00001],
freeze_layers=[-1, 0, 0])
def get_model(freeze_layers=-1, lr=0.01, output_dim=1, weights="imagenet"):
base_model = keras.applications.DenseNet201(include_top=False, weights=weights,
input_shape=model_config.image_shape, pooling="avg")
x = base_model.output
x = Dense(256, use_bias=False)(x)
x = BatchNormalization()(x)
x = Activation("relu")(x)
import keras
from keras.layers import Dense, BatchNormalization, Activation
import config
from util import data_loader
from util import keras_util
from util.keras_util import KerasModelConfig
model_config = KerasModelConfig(k_fold_file="1.txt",
model_path=os.path.abspath(__file__),
image_resolution=224,
data_type=[config.DATA_TYPE_ORIGINAL],
label_position=[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12],
label_up_sampling=[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 50],
downsampling=0.5,
train_batch_size=[16, 16, 16],
data_visualization=True,
val_batch_size=256,
predict_batch_size=256,
epoch=[1, 4, 8],
lr=[0.0005, 0.00005, 0.000005],
freeze_layers=[-1, 0.9, 0.7])
def get_model(freeze_layers=-1, lr=0.01, output_dim=1, weights="imagenet"):
base_model = keras.applications.DenseNet201(include_top=False, weights=weights,
input_shape=model_config.image_shape, pooling="avg")
x = base_model.output
x = Dense(256, use_bias=False)(x)
x = BatchNormalization()(x)
from keras.layers import Dense, BatchNormalization, Activation
import config
from util import data_loader
from util import keras_util
from util.keras_util import KerasModelConfig
model_config = KerasModelConfig(
k_fold_file="1.txt",
model_path=os.path.abspath(__file__),
image_resolution=224,
data_type=[config.DATA_TYPE_ORIGINAL],
label_position=[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12],
label_color_augment=[0, 1, 3, 5, 6, 7, 9, 10, 11, 12],
label_up_sampling=[0, 0, 0, 40, 0, 0, 0, 0, 0, 0, 0, 0, 0],
data_visualization=True,
downsampling=0.8,
train_batch_size=[32, 32, 32],
val_batch_size=256,
predict_batch_size=256,
epoch=[2, 4, 10],
lr=[0.0005, 0.00005, 0.000005],
freeze_layers=[-1, 0.6, 5],
input_norm=False)
def get_model(freeze_layers=-1, lr=0.01, output_dim=1, weights="imagenet"):
base_model = keras.applications.DenseNet169(
include_top=False,
weights=weights,
input_shape=model_config.image_shape,
import keras
from keras.layers import Dense, BatchNormalization, Activation
import config
from util import data_loader
from util import keras_util
from util import metrics
from util.keras_util import KerasModelConfig
model_config = KerasModelConfig(
k_fold_file="1.txt",
image_resolution=224,
model_path=os.path.abspath(__file__),
data_type=[config.DATA_TYPE_SEGMENTED],
label_position=[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12],
train_batch_size=[32, 32, 32],
val_batch_size=64,
predict_batch_size=128,
epoch=[1, 6, 12],
lr=[0.001, 0.0001, 0.00001],
freeze_layers=[-1, 0.5, 0])
def get_model(freeze_layers=-1, lr=0.01, output_dim=1, weights="imagenet"):
base_model = keras.applications.InceptionV3(
weights=weights,
include_top=False,
input_shape=model_config.image_shape,
pooling="max")
x = base_model.output
x = Dense(256, activation="relu", use_bias=False)(x)
from keras.layers import Dense
import config
from util import data_loader
from util import keras_util
from util import metrics
from util.keras_util import KerasModelConfig
model_config = KerasModelConfig(
k_fold_file="1.txt",
model_path=os.path.abspath(__file__),
image_resolution=360,
data_type=[config.DATA_TYPE_ORIGINAL],
label_position=[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12],
label_color_augment=[0, 1, 3, 5, 6, 7, 9, 10, 11, 12],
train_batch_size=[16, 16],
val_batch_size=128,
predict_batch_size=128,
epoch=[2, 6],
lr=[0.0001, 0.00001],
freeze_layers=[0, 0],
tta_flip=True,
data_visualization=True)
def get_model(freeze_layers=-1, lr=0.01, output_dim=1, weights="imagenet"):
base_model = keras.applications.Xception(
include_top=False,
weights=weights,
input_shape=model_config.image_shape,
pooling="avg")
import config
from util import data_loader
from util import keras_util
from util import metrics
from util.keras_util import KerasModelConfig
model_config = KerasModelConfig(
k_fold_file="1.txt",
model_path=os.path.abspath(__file__),
image_resolution=324,
data_type=[config.DATA_TYPE_ORIGINAL],
label_position=[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12],
label_color_augment=[0, 1, 3, 5, 6, 7, 9, 10, 11, 12],
train_batch_size=[16, 16],
label_up_sampling=[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 20],
data_visualization=True,
downsampling=0.8,
val_batch_size=256,
predict_batch_size=256,
epoch=[1, 6],
lr=[0.0001, 0.00001],
freeze_layers=[5, 5],
tta_crop=True)
def get_model(freeze_layers=-1, lr=0.01, output_dim=1, weights="imagenet"):
base_model = keras.applications.InceptionV3(
weights=weights,
include_top=False,
input_shape=model_config.image_shape,
import keras
from keras.layers import Dense, BatchNormalization, Activation, Flatten, GlobalAveragePooling2D
import config
from util import data_loader
from util import keras_util
from util.keras_util import KerasModelConfig
from util import path
model_config = KerasModelConfig(k_fold_file="1.txt",
model_path=os.path.abspath(__file__),
image_resolution=224,
data_type=[config.DATA_TYPE_ORIGINAL],
label_position=[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12],
label_color_augment=[0, 1, 3, 5, 6, 7, 9, 10, 11, 12],
label_up_sampling=[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 50],
downsampling=0.8,
train_batch_size=[16, 16, 16],
val_batch_size=128,
predict_batch_size=128,
epoch=[1, 4, 8],
lr=[0.001, 0.0001, 0.00001],
freeze_layers=[-1, 0.6, 5])
def get_model(freeze_layers=-1, lr=0.01, output_dim=1, weights=os.path.join(path.root_path, 'resnet152_weights_tf.h5')):
base_model = resnet152.resnet152_model(resolution=model_config.image_resolution, weights_path=weights)
base_model.layers.pop()
base_model.layers.pop()
x = GlobalAveragePooling2D()(base_model.layers[-1].output)
x = Dense(256, use_bias=False)(x)
x = BatchNormalization()(x)