config['scaling'] = "minmax"
if "np.log" in config['scaling']:
images = np.log1p(images)
# set tf random seed
tf.random.set_seed(config['random_seed'])
# ================== Import Data ==================
with tf.device(get_tf_device(20)):
padding = 'same'
model = Sequential()
model.add(Dense(4, activation='relu', input_shape=(256, )))
model.compile(
loss='mse',
optimizer='adam',
)
print(model.summary())
# Run experiment
experiment = Experiment(
model=model,
config=config,
model_type="regression",
experiment_name="generate_results_pos_double_linreg",
)
experiment.run_kfold(
images[double_indices],
normalize_position_data(positions[double_indices]),
)
experiment.save(save_model=True, save_indices=False)
print("Finished experiment:", experiment.id)
# log-scale the images if desireable
config['scaling'] = "minmax"
if "np.log" in config['scaling']:
images = np.log1p(images)
# set tf random seed
tf.random.set_seed(config['random_seed'])
# ================== Import Data ==================
with tf.device(get_tf_device(20)):
model = Sequential()
model.add(Dense(2, activation='relu', input_shape=(256, )))
model.compile(
loss='mse',
optimizer='adam',
)
print(model.summary())
# Run experiment
experiment = Experiment(
model=model,
config=config,
model_type="regression",
experiment_name="generate_results_energies_double_linreg",
)
experiment.run_kfold(
images[double_indices],
energies[double_indices],
)
experiment.save(save_model=True, save_indices=False)
print("Finished experiment:", experiment.id)
input_shape=(16, 16, 1)))
model.add(Conv2D(64, (3, 3), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Conv2D(64, (3, 3), activation='relu'))
model.add(Conv2D(64, (3, 3), activation='relu'))
model.add(Flatten())
model.add(Dense(128, activation='relu'))
model.add(Dense(1, activation='sigmoid'))
model.compile(optimizer='adam',
loss='binary_crossentropy',
metrics=['accuracy'])
# Run experiment
experiment = Experiment(model=model,
config=config,
model_type="classification",
experiment_name=search_name)
experiment.run(
normalize_image_data(images[train_idx]),
labels[train_idx],
normalize_image_data(images[val_idx]),
labels[val_idx],
)
experiment.save()
id_param[experiment.id] = {
'batch_size': b_size,
}
search_path = get_git_root() + "experiments/searches/"
with open(search_path + search_name + ".json", "w") as fp:
json.dump(id_param, fp, indent=2)
tf.random.set_seed(config['random_seed'])
# ================== Import Data ==================
with tf.device(get_tf_device(20)):
padding = 'same'
model = Sequential()
model.add(
Conv2D(8, kernel_size=3, activation='relu', input_shape=(16, 16, 1),
padding='same')
)
model.add(Flatten())
model.add(Dense(2, activation='linear'))
model.compile(
loss='mse',
optimizer='adam',
)
print(model.summary())
# Run experiment
experiment = Experiment(
model=model,
config=config,
model_type="regression",
experiment_name="generate_results_cnn_small",
)
experiment.run_kfold(
images[single_indices],
normalize_position_data(positions[single_indices])[:, :2],
)
experiment.save(save_model=True, save_indices=False)
print("Finished experiment:", experiment.id)
model.add(Conv2D(32, kernel_size=3, activation='relu', padding='same'))
model.add(Conv2D(32, kernel_size=3, activation='relu', padding='same'))
model.add(Conv2D(32, kernel_size=3, activation='relu', padding='same'))
model.add(Flatten())
model.add(Dense(1, activation='sigmoid'))
model.compile(optimizer='adam',
loss='binary_crossentropy',
metrics=['accuracy'])
models['cnn_deep'] = model
# Run experiments
for k, m in models.items():
print(m.summary())
experiment = Experiment(model=m,
config=config,
model_type="classification",
experiment_name="full_training_classifier_" +
k)
if "logistic" in k or "dense" in k:
experiment.run(
images.reshape(images.shape[0], 256),
labels,
)
else:
experiment.run(
images,
labels,
)
print("Outputting model:", k)
experiment.save()
mpath = experiment.config['path_args']['models'] + experiment.id + ".h5"
model = Sequential()
model.add(
Conv2D(8,
kernel_size=3,
activation='relu',
input_shape=images.shape[1:],
padding='same'))
model.add(Flatten())
model.add(Dense(1, activation='sigmoid'))
model.compile(optimizer='adam',
loss='binary_crossentropy',
metrics=['accuracy'])
# Run experiment
experiment = Experiment(model=model,
config=config,
model_type="classification",
experiment_name="full_training_cnn_small")
experiment.run(
images,
labels,
)
experiment.save(save_model=True, save_indices=False)
print("Finished experiment:", experiment.id)
lpath = experiment.config['path_args']['models'] + "models.log"
log = open(lpath, "a")
log.write(experiment.id + ":\n")
log.write(os.path.basename(__file__) + "\n")
log.write(repr(config) + "\n")
log.close()
if "np.log" in config['scaling']:
images = np.log1p(images)
# For the pretrained vgg model we have to use data with 3 channels.
# This is solved by concatenating the images with themselves
images = np.concatenate((images, images, images), axis=-1)
# set tf random seed
tf.random.set_seed(config['random_seed'])
with tf.device(get_tf_device(20)):
# Build model
model = pretrained_model("VGG16", input_dim=(16, 16, 3))
model.add(Dense(512, activation='relu'))
model.add(Dense(512, activation='relu'))
model.add(Dense(1, activation='sigmoid'))
model.compile(optimizer='adam',
loss='binary_crossentropy',
metrics=['accuracy'])
# Run experiment
experiment = Experiment(model=model,
config=config,
model_type="classification",
experiment_name="full_training_pretrained_vgg16")
experiment.run_kfold(
images,
labels,
)
experiment.save(save_model=True, save_indices=False)
print("Finished experiment:", experiment.id)
# set tf random seed
tf.random.set_seed(config['random_seed'])
with tf.device(get_tf_device(20)):
# Small Dense network
model = Sequential()
model.add(InputLayer(input_shape=(256, )))
model.add(Dense(8, activation='relu'))
model.add(Dense(1, activation='sigmoid'))
model.compile(optimizer='adam',
loss='binary_crossentropy',
metrics=['accuracy'])
# Run experiment
experiment = Experiment(model=model,
config=config,
model_type="classification",
experiment_name="full_training_dense_small")
experiment.run_kfold(
images.reshape(images.shape[0], 256),
labels,
)
experiment.save(save_model=True, save_indices=False)
print("Finished experiment:", experiment.id)
lpath = experiment.config['path_args']['models'] + "models.log"
log = open(lpath, "a")
log.write(experiment.id + ":\n")
log.write(os.path.basename(__file__) + "\n")
log.write(repr(config) + "\n")
log.close()
outputs = DSNT()(x)
model = tf.keras.Model(inputs=inputs, outputs=outputs[1])
prediction_model = tf.keras.Model(inputs=inputs, outputs=outputs)
model.compile(
optimizer='adam',
loss='mse',
)
prediction_model.compile(
optimizer='adam',
loss='mse',
)
print(model.summary())
# Run experiment
experiment = Experiment(model=model,
config=config,
model_type="regression",
experiment_name=search_name)
experiment.run(
normalize_image_data(images[single_indices]),
normalize_position_data(positions[single_indices])[:, :2],
)
experiment.save()
mpath = experiment.config['path_args']['models'] + experiment.id + ".h5"
prediction_model.save(mpath)
heatmaps, coords = prediction_model.predict(
images[single_indices][experiment.indices['fold_0']['val_idx']])
np.save("dsnt_heatmaps_pred.npy", heatmaps)
np.save("dsnt_coords_pred.npy", coords)
print("Finished experiment.")
print("Name:", search_name)
print("id:", experiment.id)
model.add(Conv2D(32, kernel_size=(3, 3),
activation='relu', input_shape=(16, 16, 1),
padding=padding))
model.add(Conv2D(64, (3, 3), activation='relu', padding=padding))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Conv2D(64, (3, 3), activation='relu', padding=padding))
model.add(Conv2D(64, (3, 3), activation='relu', padding=padding))
model.add(Flatten())
model.add(Dense(256, activation='relu'))
model.add(Dense(1, activation='linear'))
model.compile(
loss='mse',
optimizer='adam',
)
print(model.summary())
# Run experiment
experiment = Experiment(
model=model,
config=config,
model_type="regression",
experiment_name=search_name
)
experiment.run(
normalize_image_data(images[single_indices]),
energies[single_indices, 0],
)
experiment.save()
mpath = experiment.config['path_args']['models'] + experiment.id + ".h5"
model.save(mpath)
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Conv2D(64, (3, 3), activation='relu', padding=padding))
model.add(Conv2D(64, (3, 3), activation='relu', padding=padding))
model.add(Flatten())
model.add(Dense(128, activation='relu'))
model.add(Dense(1, activation='sigmoid'))
model.compile(
optimizer='adam',
loss='binary_crossentropy',
metrics=['accuracy']
)
# Run experiment
experiment = Experiment(
model=model,
config=config,
model_type="classification",
experiment_name="full_training_classifier"
)
experiment.run(
images,
labels,
)
experiment.save()
mpath = experiment.config['path_args']['models'] + experiment.id + ".h5"
model.save(mpath)
print(experiment.id)
lpath = experiment.config['path_args']['models'] + "models.log"
log = open(lpath,"a")
log.write(experiment.id + ":\n")
log.write(os.path.basename(__file__)+"\n")
log.write(repr(config) + "\n")
