def get_loss(root, split, net, recon_wei, choice):
if choice == 'w_bce':
pos_class_weight = load_class_weights(root=root, split=split)
loss = weighted_binary_crossentropy_loss(pos_class_weight)
elif choice == 'bce':
loss = 'binary_crossentropy'
elif choice == 'dice':
loss = dice_loss
elif choice == 'multi_dice':
loss = multiclass_dice_loss
elif choice == 'w_mar':
pos_class_weight = load_class_weights(root=root, split=split)
loss = margin_loss(margin=0.4,
downweight=0.5,
pos_weight=pos_class_weight)
elif choice == 'mar':
loss = margin_loss(margin=0.4, downweight=0.5, pos_weight=1.0)
else:
raise Exception("Unknow loss_type")
if net.find('caps') != -1:
return {
'out_seg': loss,
'out_recon': 'mse'
}, {
'out_seg': 1.,
'out_recon': recon_wei
}
else:
return loss, None
def get_loss(root, split, net, choice):
if choice == 'w_bce':
pos_class_weight = load_class_weights(root=root, split=split)
loss = weighted_binary_crossentropy_loss(pos_class_weight)
elif choice == 'bce':
loss = 'binary_crossentropy'
elif choice == 'dice':
loss = dice_loss
elif choice == 'w_mar':
pos_class_weight = load_class_weights(root=root, split=split)
loss = margin_loss(margin=0.4,
downweight=0.5,
pos_weight=pos_class_weight)
elif choice == 'mar':
loss = margin_loss(margin=0.4, downweight=0.5, pos_weight=1.0)
else:
raise Exception("Unknow loss_type")
return loss, None
def get_loss(root, split, net, recon_wei, choice):
if choice == 'w_bce':
pos_class_weight = load_class_weights(root=root, split=split)
loss = weighted_binary_crossentropy_loss(pos_class_weight)
elif choice == 'bce':
loss = 'binary_crossentropy'
elif choice == 'dice':
loss = weighted_dice_loss(S_PRESENCE)
elif choice == 'w_mar':
pos_class_weight = load_class_weights(root=root, split=split)
loss = margin_loss(margin=0.4,
downweight=0.5,
pos_weight=pos_class_weight)
elif choice == 'mar':
loss = margin_loss(margin=0.4, downweight=0.5, pos_weight=1.0)
elif choice == 'cce':
loss = 'categorical_crossentropy'
elif choice == 'scce':
loss = 'sparse_categorical_crossentropy'
elif choice == 'spread':
loss = spread_loss(epoch_step=EpochCounter.counter)
elif choice == 'w_spread':
if 'Spectralis' in root:
weights = np.array(S_PRESENCES)
else:
weights = np.array(C_PRESENCE)
loss = weighted_spread_loss(weights=weights,
epoch_step=EpochCounter.counter)
else:
raise Exception("Unknow loss_type")
if net.find('caps') != -1:
# return {'out_seg': loss, 'recon0': weighted_mse_loss(S_PIXEL_MSE[0]),
# 'recon1': weighted_mse_loss(S_PIXEL_MSE[1]),
# 'recon2': weighted_mse_loss(S_PIXEL_MSE[2]),
# 'recon3': weighted_mse_loss(S_PIXEL_MSE[3])}, {'out_seg': 1., 'recon0': recon_wei,
# 'recon1': recon_wei,
# 'recon2': recon_wei,
# 'recon3': recon_wei}
return {'out_seg': loss}, None
else:
return loss, None
def get_loss(root, split, net, recon_wei, choice):
if choice == 'w_bce':
pos_class_weight = load_class_weights(root=root, split=split)
loss = weighted_binary_crossentropy_loss(pos_class_weight)
elif choice == 'bce':
loss = 'binary_crossentropy'
elif choice == 'dice':
loss = dice_loss
elif choice == 'w_mar':
pos_class_weight = load_class_weights(root=root, split=split)
loss = margin_loss(margin=0.4, downweight=0.5, pos_weight=pos_class_weight)
elif choice == 'mar':
loss = margin_loss(margin=0.4, downweight=0.5, pos_weight=1.0)
else:
raise Exception("Unknow loss_type")
if net.find('caps') != -1:
return {'out_seg': loss, 'out_recon': 'mse'}, {'out_seg': 1., 'out_recon': recon_wei}
else:
return loss, None
def get_loss(net, recon_wei, attr_wei, choice, classes):
if choice == 'ce':
if classes == 1:
loss = 'binary_crossentropy'
else:
loss = 'categorical_crossentropy'
attr_loss = 'binary_crossentropy'
elif choice == 'mar':
loss, attr_loss = margin_loss(margin=0.4,
downweight=0.5,
pos_weight=1.0), margin_loss(
margin=0.4,
downweight=0.5,
pos_weight=1.0)
elif choice == 'mse':
loss, attr_loss = 'mse', 'mse'
elif choice == 'kl':
if classes == 1:
raise Exception(
"Cannot compute KL divergence with scalar regression output.")
loss = tf.keras.losses.KLDivergence()
attr_loss = 'mse'
else:
raise Exception("Unknown loss_type")
if net.find('capsnet') != -1:
return {
'out_mal': loss,
'out_recon': 'mse'
}, {
'out_mal': 1.,
'out_recon': recon_wei
}
elif net.find('xcaps') != -1:
return {'out_mal': loss, 'out_recon': 'mse', 'out_attr_0': attr_loss, 'out_attr_1': attr_loss,
'out_attr_2': attr_loss, 'out_attr_3': attr_loss, 'out_attr_4': attr_loss, 'out_attr_5': attr_loss}, \
{'out_mal': 1., 'out_attr_0': attr_wei, 'out_attr_1': attr_wei, 'out_attr_2': attr_wei,
'out_attr_3': attr_wei, 'out_attr_4': attr_wei, 'out_attr_5': attr_wei, 'out_recon': recon_wei}
else:
return loss, None
def get_loss(train_list, split, net, recon_wei, choice):
if choice == 'dice':
loss = dice_loss
elif choice == 'mar':
loss = margin_loss(margin=0.4, downweight=0.5, pos_weight=1.0)
elif choice == 'bce_dice':
loss = bce_dice_loss
elif choice == 'bce':
loss = 'binary_crossentropy'
else:
raise Exception("Unknown loss_type")
if net.find('caps') != -1:
return {
'out_seg': loss,
'out_recon': 'mse'
}, {
'out_seg': 1.,
'out_recon': recon_wei
}
else:
return loss, None