x_test = np.array(x_test)
#*******************************************
# Define traing params
#*******************************************
learning_rate = 0.00002
batch_size = 6
display_step = 1
epochs = 100
class_weights = [1, 1, 12]
generator = BalanceCovidDataset(data_dir=data_path,
csv_file=train_csv,
covid_percent=0.3,
class_weights=class_weights,
batch_size=batch_size)
total_batch = len(generator)
model = keras_model_build()
opt = tf.keras.optimizers.Adam(lr=learning_rate)
# ********************************************
# Train model
#*********************************************
model.compile(loss='categorical_crossentropy',
optimizer=opt,
metrics=['accuracy'])
hist = model.fit(generator,
# For detection of no pneumonia/non-COVID-19 pneumonia/COVID-19 pneumonia
mapping = {
'normal': 0,
'pneumonia': 1,
'COVID-19': 2
}
class_weights = [1., 1., args.covid_weight]
else:
raise Exception('''COVID-Net currently only supports 2 class COVID-19 positive/negative detection
or 3 class detection of no pneumonia/non-COVID-19 pneumonia/COVID-19 pneumonia''')
generator = BalanceCovidDataset(data_dir=args.datadir,
csv_file=args.trainfile,
batch_size=batch_size,
input_shape=(args.input_size, args.input_size),
n_classes=args.n_classes,
mapping=mapping,
covid_percent=args.covid_percent,
class_weights=class_weights,
top_percent=args.top_percent)
with tf.Session() as sess:
tf.get_default_graph()
saver = tf.train.import_meta_graph(os.path.join(args.weightspath, args.metaname))
graph = tf.get_default_graph()
image_tensor = graph.get_tensor_by_name(args.in_tensorname)
labels_tensor = graph.get_tensor_by_name(args.label_tensorname)
sample_weights = graph.get_tensor_by_name(args.weights_tensorname)
pred_tensor = graph.get_tensor_by_name(args.logit_tensorname)
# output path
outputPath = './output/'
runID = args.name + '-lr' + str(learning_rate)
runPath = outputPath + runID
pathlib.Path(runPath).mkdir(parents=True, exist_ok=True)
print('Output: ' + runPath)
with open(args.trainfile) as f:
trainfiles = f.readlines()
with open(args.testfile) as f:
testfiles = f.readlines()
generator = BalanceCovidDataset(data_dir=args.datadir,
csv_file=args.trainfile,
batch_size=batch_size,
input_shape=(args.input_size, args.input_size),
covid_percent=args.covid_percent,
class_weights=[1., 1., args.covid_weight],
top_percent=args.top_percent)
with tf.Session() as sess:
tf.get_default_graph()
saver = tf.train.import_meta_graph(os.path.join(args.weightspath, args.metaname))
graph = tf.get_default_graph()
image_tensor = graph.get_tensor_by_name(args.in_tensorname)
labels_tensor = graph.get_tensor_by_name(args.label_tensorname)
sample_weights = graph.get_tensor_by_name(args.weights_tensorname)
pred_tensor = graph.get_tensor_by_name(args.logit_tensorname)
# loss expects unscaled logits since it performs a softmax on logits internally for efficiency
def train_tf(epochs, lr, bs, weightspath, metaname, ckptname, trainfile,
testfile, name, datadir, covid_weight, covid_percent, input_size,
top_percent, in_tensorname, out_tensorname, logit_tensorname,
label_tensorname, weights_tensorname, input_data_dir,
output_data_dir):
# Parameters
learning_rate = lr
batch_size = bs
display_step = 1
# output path
outputPath = output_data_dir
runID = name + '-lr' + str(learning_rate)
runPath = outputPath + runID
pathlib.Path(runPath).mkdir(parents=True, exist_ok=True)
print('Output: ' + runPath)
with open(trainfile) as f:
trainfiles = f.readlines()
with open(testfile) as f:
testfiles = f.readlines()
generator = BalanceCovidDataset(data_dir=datadir,
csv_file=trainfile,
batch_size=batch_size,
input_shape=(input_size, input_size),
covid_percent=covid_percent,
class_weights=[1., 1., covid_weight],
top_percent=top_percent)
with tf.Session() as sess:
tf.get_default_graph()
saver = tf.train.import_meta_graph(os.path.join(weightspath, metaname))
graph = tf.get_default_graph()
image_tensor = graph.get_tensor_by_name(in_tensorname)
labels_tensor = graph.get_tensor_by_name(label_tensorname)
sample_weights = graph.get_tensor_by_name(weights_tensorname)
pred_tensor = graph.get_tensor_by_name(logit_tensorname)
# loss expects unscaled logits since it performs a softmax on logits internally for efficiency
# Define loss and optimizer
loss_op = tf.reduce_mean(
tf.nn.softmax_cross_entropy_with_logits_v2(
logits=pred_tensor, labels=labels_tensor) * sample_weights)
optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate)
train_op = optimizer.minimize(loss_op)
# Initialize the variables
init = tf.global_variables_initializer()
# Run the initializer
sess.run(init)
# load weights
saver.restore(sess, os.path.join(weightspath, ckptname))
#saver.restore(sess, tf.train.latest_checkpoint(weightspath))
# save base model
saver.save(sess, os.path.join(runPath, 'model'))
print('Saved baseline checkpoint')
print('Baseline eval:')
eval(sess, graph, testfiles, os.path.join(datadir, 'test'),
in_tensorname, out_tensorname, input_size)
# Training cycle
print('Training started')
total_batch = len(generator)
progbar = tf.keras.utils.Progbar(total_batch)
for epoch in range(epochs):
for i in range(total_batch):
# Run optimization
batch_x, batch_y, weights = next(generator)
sess.run(train_op,
feed_dict={
image_tensor: batch_x,
labels_tensor: batch_y,
sample_weights: weights
})
progbar.update(i + 1)
if epoch % display_step == 0:
pred = sess.run(pred_tensor, feed_dict={image_tensor: batch_x})
loss = sess.run(loss_op,
feed_dict={
pred_tensor: pred,
labels_tensor: batch_y,
sample_weights: weights
})
print("Epoch:", '%04d' % (epoch + 1), "Minibatch loss=",
"{:.9f}".format(loss))
eval(sess, graph, testfiles, os.path.join(datadir, 'test'),
in_tensorname, out_tensorname, input_size)
saver.save(sess,
os.path.join(runPath, 'model'),
global_step=epoch + 1,
write_meta_graph=False)
print('Saving checkpoint at epoch {}'.format(epoch + 1))
print("Optimization Finished!")
display_step = 1
# output path
outputPath = './output/'
runID = args.name + '-lr' + str(learning_rate)
runPath = outputPath + runID
pathlib.Path(runPath).mkdir(parents=True, exist_ok=True)
print('Output: ' + runPath)
with open(args.trainfile) as f:
trainfiles = f.readlines()
with open(args.testfile) as f:
testfiles = f.readlines()
generator = BalanceCovidDataset(data_dir=args.datadir,
csv_file=args.trainfile,
covid_percent=args.covid_percent,
class_weights=[1., 1., args.covid_weight])
with tf.Session() as sess:
tf.get_default_graph()
saver = tf.train.import_meta_graph(
os.path.join(args.weightspath, args.metaname))
graph = tf.get_default_graph()
image_tensor = graph.get_tensor_by_name("input_1:0")
labels_tensor = graph.get_tensor_by_name("dense_3_target:0")
sample_weights = graph.get_tensor_by_name("dense_3_sample_weights:0")
pred_tensor = graph.get_tensor_by_name("dense_3/MatMul:0")
# loss expects unscaled logits since it performs a softmax on logits internally for efficiency
default='norm_dense_1_sample_weights:0',
type=str,
help='Name of sample weights tensor for loss')
args = parser.parse_args()
# Parameters
learning_rate = args.lr
batch_size = args.bs
display_step = 1
generator_test = BalanceCovidDataset(data_dir=args.datadir,
csv_file=args.testfile,
is_training=False,
batch_size=1,
input_shape=(args.input_size,
args.input_size),
covid_percent=args.covid_percent,
class_weights=[1., 1., 1.],
top_percent=args.top_percent)
model = keras_model_build()
optimizer = tf.keras.optimizers.Adam(lr=learning_rate)
model.compile(loss='categorical_crossentropy',
optimizer=optimizer,
metrics=['accuracy'
]) # loss nie jest poprawiony i metryka jeszcze
model.load_weights("./checkpoints_douczany_2/model.20/")