def train(self, class_weight=None, submodel=None):
"""Train a model"""
callback_list = callbacks.create(self.log_dir)
if submodel == "spatial":
#The spatial model is very shallow compared to spectral, train for longer
self.spatial_model.fit(self.train_split,
epochs=int(self.config["train"]["epochs"] /
2),
validation_data=self.val_split,
callbacks=callback_list,
class_weight=class_weight)
elif submodel == "spectral":
#one for each loss layer
self.spectral_model.fit(self.train_split,
epochs=int(self.config["train"]["epochs"] /
2),
validation_data=self.val_split,
callbacks=callback_list,
class_weight=class_weight)
else:
self.model.fit(self.train_split,
epochs=self.config["train"]["epochs"],
validation_data=self.val_split,
callbacks=callback_list,
class_weight=class_weight)
def ensemble(self, experiment, class_weight=None):
if self.val_split is None:
print("Cannot run callbacks without validation data, skipping...")
callback_list = None
label_names = None
elif experiment is None:
print("Cannot run callbacks without comet experiment, skipping...")
callback_list = None
label_names = None
else:
if self.classes_file is not None:
labeldf = pd.read_csv(self.classes_file)
label_names = list(labeldf.taxonID.values)
else:
label_names = None
callback_list = callbacks.create(log_dir=self.log_dir,
experiment=experiment,
validation_data=self.val_split,
train_data=self.train_split,
label_names=label_names,
train_shp=self.train_shp,
submodel="ensemble")
#Train ensemble layer
self.ensemble_model.fit(
self.train_split,
epochs=self.config["train"]["ensemble"]["epochs"],
validation_data=self.val_split,
callbacks=callback_list,
class_weight=class_weight)
def train(self, class_weight=None):
"""Train a model"""
callback_list = callbacks.create()
self.model.fit(
self.train_split,
epochs=self.config["train"]["epochs"],
validation_data=self.val_split,
callbacks=callback_list,
class_weight=class_weight
)
def ensemble(self, experiment, class_weight=None, freeze = True, train=True):
#Manually override batch size
self.config["train"]["batch_size"] = self.config["train"]["ensemble"]["batch_size"]
self.read_data(mode="ensemble")
self.ensemble_model = Hang.ensemble([self.HSI_model, self.RGB_model], freeze=freeze, classes=self.classes)
if train:
self.ensemble_model.compile(
loss="categorical_crossentropy",
optimizer=tf.keras.optimizers.Adam(
lr=float(self.config["train"]["learning_rate"])),
metrics=[tf.keras.metrics.CategoricalAccuracy(
name='acc')])
if self.val_split is None:
print("Cannot run callbacks without validation data, skipping...")
callback_list = None
label_names = None
elif experiment is None:
print("Cannot run callbacks without comet experiment, skipping...")
callback_list = None
label_names = None
else:
if self.classes_file is not None:
labeldf = pd.read_csv(self.classes_file)
label_names = list(labeldf.taxonID.values)
else:
label_names = None
callback_list = callbacks.create(log_dir=self.log_dir,
experiment=experiment,
validation_data=self.val_split,
train_data=self.train_split,
label_names=label_names,
submodel="ensemble")
#Train ensemble layer
self.ensemble_model.fit(
self.train_split,
epochs=self.config["train"]["epochs"],
validation_data=self.val_split,
callbacks=callback_list,
class_weight=class_weight)
def train(self, experiment=None, class_weight=None, submodel=None):
"""Train a model with callbacks"""
if self.val_split is None:
print("Cannot run callbacks without validation data, skipping...")
callback_list = None
elif experiment is None:
print("Cannot run callbacks without comet experiment, skipping...")
callback_list = None
else:
labeldf = pd.read_csv(self.classes_file)
callback_list = callbacks.create(log_dir=self.log_dir,
experiment=experiment,
validation_data=self.val_split,
label_names=list(
labeldf.taxonID.values))
if submodel == "spatial":
#The spatial model is very shallow compared to spectral, train for longer
self.spatial_model.fit(self.train_split,
epochs=int(self.config["train"]["epochs"] /
2),
validation_data=self.val_split,
callbacks=callback_list,
class_weight=class_weight)
elif submodel == "spectral":
#one for each loss layer
self.spectral_model.fit(self.train_split,
epochs=int(self.config["train"]["epochs"] /
2),
validation_data=self.val_split,
callbacks=callback_list,
class_weight=class_weight)
else:
self.model.fit(self.train_split,
epochs=self.config["train"]["epochs"],
validation_data=self.val_split,
callbacks=callback_list,
class_weight=class_weight)
def train(self, experiment=None, class_weight=None, submodel=None, sensor="hyperspectral"):
"""Train a model with callbacks"""
if self.val_split is None:
print("Cannot run callbacks without validation data, skipping...")
callback_list = None
elif experiment is None:
print("Cannot run callbacks without comet experiment, skipping...")
callback_list = None
else:
if self.classes_file is not None:
labeldf = pd.read_csv(self.classes_file)
label_names = list(labeldf.taxonID.values)
else:
label_names = None
callback_list = callbacks.create(log_dir=self.log_dir,
experiment=experiment,
validation_data=self.val_split,
train_data=self.train_split,
label_names=label_names,
submodel=submodel)
if submodel == "spatial":
if sensor == "hyperspectral":
self.HSI_spatial.fit(self.train_split,
epochs=int(self.config["train"]["epochs"]),
validation_data=self.val_split,
callbacks=callback_list,
class_weight=class_weight)
elif sensor == "RGB":
self.RGB_spatial.fit(self.train_split,
epochs=int(self.config["train"]["epochs"]),
validation_data=self.val_split,
callbacks=callback_list,
class_weight=class_weight)
elif submodel == "spectral":
if sensor == "hyperspectral":
self.HSI_spectral.fit(self.train_split,
epochs=int(self.config["train"]["epochs"]),
validation_data=self.val_split,
callbacks=callback_list,
class_weight=class_weight)
elif sensor == "RGB":
self.RGB_spectral.fit(self.train_split,
epochs=int(self.config["train"]["epochs"]),
validation_data=self.val_split,
callbacks=callback_list,
class_weight=class_weight)
else:
if sensor == "hyperspectral":
self.HSI_model.fit(self.train_split,
epochs=self.config["train"]["epochs"],
validation_data=self.val_split,
callbacks=callback_list,
class_weight=class_weight)
elif sensor == "RGB":
self.RGB_model.fit(
self.train_split,
epochs=self.config["train"]["epochs"],
validation_data=self.val_split,
callbacks=callback_list,
class_weight=class_weight)
def ensemble(self, experiment, class_weight=None, freeze=True, train=True):
self.classes = pd.read_csv(self.classes_file).shape[0]
self.read_data(mode="ensemble")
if self.val_split is None:
print("Cannot run callbacks without validation data, skipping...")
callback_list = None
label_names = None
elif experiment is None:
print("Cannot run callbacks without comet experiment, skipping...")
callback_list = None
label_names = None
else:
if self.classes_file is not None:
labeldf = pd.read_csv(self.classes_file)
label_names = list(labeldf.taxonID.values)
else:
label_names = None
callback_list = callbacks.create(log_dir=self.log_dir,
experiment=experiment,
validation_data=self.val_split,
train_data=self.train_split,
label_names=label_names,
train_shp=self.train_shp,
submodel="ensemble")
print("callback list is {}".format(callback_list))
if self.config["train"]["gpus"] > 1:
with self.strategy.scope():
self.ensemble_model = Hang.learned_ensemble(
HSI_model=self.HSI_model,
metadata_model=self.metadata_model,
freeze=freeze,
classes=self.classes)
if train:
self.ensemble_model.compile(
loss="categorical_crossentropy",
optimizer=tf.keras.optimizers.Adam(
lr=float(self.config["train"]["learning_rate"])),
metrics=[
tf.keras.metrics.CategoricalAccuracy(name='acc')
])
#Train ensemble layer
self.ensemble_model.fit(
self.train_split,
epochs=self.config["train"]["ensemble"]["epochs"],
validation_data=self.val_split,
callbacks=callback_list,
class_weight=class_weight)
else:
self.ensemble_model = Hang.learned_ensemble(
HSI_model=self.HSI_model,
metadata_model=self.metadata_model,
freeze=freeze,
classes=self.classes)
if train:
self.ensemble_model.compile(
loss="categorical_crossentropy",
optimizer=tf.keras.optimizers.Adam(
lr=float(self.config["train"]["learning_rate"])),
metrics=[tf.keras.metrics.CategoricalAccuracy(name='acc')])
#Train ensemble layer
self.ensemble_model.fit(
self.train_split,
epochs=self.config["train"]["ensemble"]["epochs"],
validation_data=self.val_split,
callbacks=callback_list,
class_weight=class_weight)
config="/home/b.weinstein/DeepTreeAttention/conf/tree_config.yml",
log_dir=save_dir)
model.read_data("HSI")
model.create()
baseline = vanilla.create(
height=model.config["train"]["HSI"]["crop_size"],
width=model.config["train"]["HSI"]["crop_size"],
channels=model.config["train"]["HSI"]["sensor_channels"],
classes=model.classes)
baseline.compile(loss="categorical_crossentropy",
optimizer=tf.keras.optimizers.Adam(
lr=float(model.config["train"]["learning_rate"])),
metrics=[tf.keras.metrics.CategoricalAccuracy(name='acc')])
labeldf = pd.read_csv(model.classes_file)
label_names = list(labeldf.taxonID.values)
callback_list = callbacks.create(experiment=experiment,
train_data=model.train_split,
validation_data=model.val_split,
train_shp=model.train_shp,
log_dir=None,
label_names=label_names,
submodel=False)
baseline.fit(model.train_split,
epochs=model.config["train"]["ensemble"]["epochs"],
validation_data=model.val_split,
callbacks=callback_list)
