diff --git a/pysegcnn/main/config.py b/pysegcnn/main/config.py
index 8780589566612c47393c92e1f2db1a2b81492ba9..c307f1db46eae24f75bd0fda872421e803a36cd5 100644
--- a/pysegcnn/main/config.py
+++ b/pysegcnn/main/config.py
@@ -30,6 +30,12 @@ DATASET_PATH = os.path.join(DRIVE_PATH, DATASET_NAME)
# DATASET_PATH = os.path.join(DRIVE_PATH, DATASET_NAME, 'Training')
# DATASET_PATH = os.path.join(DRIVE_PATH, 'ProSnow', DATASET_NAME)
+# path to store the model states
+MODEL_PATH = os.path.join(HERE, '_models/')
+
+# path to store model logs
+LOG_PATH = os.path.join(HERE, '_logs/')
+
# the dataset configuration dictionary
dataset_config = {
@@ -189,9 +195,6 @@ model_config = {
'dilation': 1 # the field of view of the kernel
},
- # path to save trained models
- 'state_path': os.path.join(HERE, '_models/'),
-
# Transfer learning -------------------------------------------------------
# Use pretrained=True only if you wish to fine-tune a pre-trained model
@@ -203,7 +206,7 @@ model_config = {
# was trained on
# whether to use a pretrained model for transfer learning
- 'transfer': True,
+ 'transfer': False,
# name of the pretrained model to apply to a different dataset
'pretrained_model': 'UNet_SparcsDataset_t125_b64_rgbn.pt',
@@ -226,6 +229,8 @@ model_config = {
# -------------------------------------------------------------------------
# whether to save the model state to disk
+ # model states are saved in: pysegcnn/main/_models
+ # model log files are saved in: pysegcnn/main/_logs
'save': True,
# whether to early stop training if the accuracy on the validation set
@@ -260,6 +265,10 @@ eval_config = {
# these options are only used for evaluating a trained model using
# pysegcnn.main.eval.py
+ # the model to evaluate
+ 'state_file': os.path.join(MODEL_PATH,
+ 'UNet_SparcsDataset_Adam_SceneSplit_s0_t005v05_t125_b64_r4g3b2n5.pt'),
+
# the dataset to evaluate the model on
# test=False, 0 means evaluating on the validation set
# test=True, 1 means evaluating on the test set
@@ -267,6 +276,7 @@ eval_config = {
'test': False,
# whether to compute and plot the confusion matrix
+ # output path is: pysegcnn/main/_graphics/
'cm': True,
# whether to predict each sample or each scene individually
@@ -299,9 +309,3 @@ eval_config = {
'alpha': 5
}
-
-# the complete configuration
-config = {**dataset_config,
- **split_config,
- **model_config,
- **eval_config}
diff --git a/pysegcnn/main/eval.py b/pysegcnn/main/eval.py
index 82fe4e37e2c9444594701187c75aacda2426842c..687c5bb487776898d20b83c9e9db6ab0c47be0e5 100644
--- a/pysegcnn/main/eval.py
+++ b/pysegcnn/main/eval.py
@@ -5,52 +5,41 @@ Created on Wed Jul 29 15:57:01 2020
@author: Daniel
"""
# builtins
-import os
+from logging.config import dictConfig
# locals
-from pysegcnn.core.trainer import (DatasetConfig, SplitConfig, ModelConfig,
- StateConfig, EvalConfig)
+from pysegcnn.core.models import Network
+from pysegcnn.core.trainer import EvalConfig, LogConfig
from pysegcnn.core.predict import predict_samples, predict_scenes
-from pysegcnn.main.config import (dataset_config, split_config, model_config,
- train_config, eval_config, HERE)
+from pysegcnn.core.logging import log_conf
from pysegcnn.core.graphics import plot_confusion_matrix, plot_loss
+from pysegcnn.main.config import eval_config
if __name__ == '__main__':
- # (i) instanciate the configurations
- dc = DatasetConfig(**dataset_config)
- sc = SplitConfig(**split_config)
- mc = ModelConfig(**model_config)
+ # instanciate the evaluation configuration
ec = EvalConfig(**eval_config)
- # (ii) instanciate the dataset
- ds = dc.init_dataset()
+ # initialize logging
+ log = LogConfig(ec.state_file)
+ dictConfig(log_conf(log.log_file))
- # (iii) instanciate the training, validation and test datasets
- train_ds, valid_ds, test_ds = sc.train_val_test_split(ds)
-
- # (iv) instanciate the model state
- state = StateConfig(ds, sc, mc)
- state_file, loss_state = state.init_state()
-
- # (vii) load pretrained model weights
- model, _ = mc.load_pretrained(state_file)
- model.state_file = state_file
+ # load the model state
+ model, _, model_state = Network.load(ec.state_file)
# plot loss and accuracy
- plot_loss(loss_state, outpath=os.path.join(HERE, '_graphics/'))
+ plot_loss(ec.state_file, outpath=ec.models_path)
# check whether to evaluate the model on the training set, validation set
# or the test set
if ec.test is None:
- ds = train_ds
+ ds = model_state['train_ds']
else:
- ds = test_ds if ec.test else valid_ds
+ ds = model_state['test_ds'] if ec.test else model_state['valid_ds']
# keyword arguments for plotting
kwargs = {'bands': ec.plot_bands,
- 'outpath': os.path.join(HERE, '_scenes/'),
'alpha': ec.alpha,
'figsize': ec.figsize}
@@ -58,16 +47,17 @@ if __name__ == '__main__':
if ec.predict_scene:
# reconstruct and predict the scenes in the validation/test set
scenes, cm = predict_scenes(ds, model, scene_id=None, cm=ec.cm,
- plot=ec.plot_scenes, **kwargs)
+ plot=ec.plot_scenes,
+ outpath=ec.scenes_path, **kwargs)
else:
# predict the samples in the validation/test set
samples, cm = predict_samples(ds, model, cm=ec.cm,
- plot=ec.plot_samples, **kwargs)
+ plot=ec.plot_samples,
+ outpath=ec.sample_path, **kwargs)
# whether to plot the confusion matrix
if ec.cm:
plot_confusion_matrix(cm, ds.dataset.labels,
- state=state_file.name.replace('.pt', '.png'),
- outpath=os.path.join(HERE, '_graphics/')
- )
+ state=ec.state_file.name.replace('.pt', '.png'),
+ outpath=ec.models_path)
diff --git a/pysegcnn/main/train.py b/pysegcnn/main/train.py
index bcffb54dd38220e5e308df477dc7428fe65db614..e28f55f0b18c8addaaceda704bda5d6c8ebcf8b1 100644
--- a/pysegcnn/main/train.py
+++ b/pysegcnn/main/train.py
@@ -6,13 +6,14 @@ Created on Tue Jun 30 09:33:38 2020
@author: Daniel
"""
# builtins
-import logging
+from logging.config import dictConfig
# locals
from pysegcnn.core.trainer import (DatasetConfig, SplitConfig, ModelConfig,
- StateConfig, NetworkTrainer)
+ StateConfig, LogConfig, NetworkTrainer)
from pysegcnn.core.logging import log_conf
-from pysegcnn.main.config import (dataset_config, split_config, model_config)
+from pysegcnn.main.config import (dataset_config, split_config, model_config,
+ LOG_PATH)
if __name__ == '__main__':
@@ -26,49 +27,42 @@ if __name__ == '__main__':
# (ii) instanciate the dataset
ds = dc.init_dataset()
- ds
# (iii) instanciate the model state
state = StateConfig(ds, sc, mc)
- state_file, loss_state = state.init_state()
+ state_file = state.init_state()
- # initialize logging
- log_file = str(state_file).replace('.pt', '_train.log')
- logging.config.dictConfig(log_conf(log_file))
+ # (iv) initialize logging
+ log = LogConfig(state_file)
+ dictConfig(log_conf(log.log_file))
- # (iv) instanciate the training, validation and test datasets and
+ # (v) instanciate the training, validation and test datasets and
# dataloaders
train_ds, valid_ds, test_ds = sc.train_val_test_split(ds)
train_dl, valid_dl, test_dl = sc.dataloaders(
train_ds, valid_ds, test_ds, batch_size=mc.batch_size, shuffle=True,
drop_last=False)
- # (iv) instanciate the model
- model = mc.init_model(ds)
+ # (vi) instanciate the model
+ model, optimizer, checkpoint_state = mc.init_model(ds, state_file)
- # (vi) instanciate the optimizer and the loss function
- optimizer = mc.init_optimizer(model)
+ # (vii) instanciate the loss function
loss_function = mc.init_loss_function()
- # (vii) resume training from an existing model checkpoint
- (model, optimizer, checkpoint_state, max_accuracy) = mc.from_checkpoint(
- model, optimizer, state_file, loss_state)
-
- # (viii) initialize network trainer class for eays model training
+ # (viii) initialize network trainer class for easy model training
trainer = NetworkTrainer(model=model,
optimizer=optimizer,
loss_function=loss_function,
train_dl=train_dl,
valid_dl=valid_dl,
+ test_dl=test_dl,
state_file=state_file,
- loss_state=loss_state,
epochs=mc.epochs,
nthreads=mc.nthreads,
early_stop=mc.early_stop,
mode=mc.mode,
delta=mc.delta,
patience=mc.patience,
- max_accuracy=max_accuracy,
checkpoint_state=checkpoint_state,
save=mc.save
)