This notebook demonstrates training a CARE model for a combined denoising and upsampling task, assuming that training data was already generated via 1_datagen.ipynb and has been saved to disk to the file data/my_training_data.npz
. Note that the training approach is exactly the same as in the standard CARE approach, what differs is the training data generation and prediction.
Note that training a neural network for actual use should be done on more (representative) data and with more training time.
More documentation is available at http://csbdeep.bioimagecomputing.com/doc/.
from __future__ import print_function, unicode_literals, absolute_import, division
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline
%config InlineBackend.figure_format = 'retina'
from tifffile import imread
from csbdeep.utils import axes_dict, plot_some, plot_history
from csbdeep.utils.tf import limit_gpu_memory
from csbdeep.io import load_training_data
from csbdeep.models import Config, UpsamplingCARE
The TensorFlow backend uses all available GPU memory by default, hence it can be useful to limit it:
# limit_gpu_memory(fraction=1/2)
(X,Y), (X_val,Y_val), axes = load_training_data('data/my_training_data.npz', validation_split=0.1, verbose=True)
c = axes_dict(axes)['C']
n_channel_in, n_channel_out = X.shape[c], Y.shape[c]
number of training images: 1382 number of validation images: 154 image size (3D): (32, 64, 64) axes: SZYXC channels in / out: 1 / 1
plt.figure(figsize=(12,3))
plot_some(X_val[:5,...,0,0],Y_val[:5,...,0,0])
plt.suptitle('5 example validation patches (ZY slice, top row: source, bottom row: target)');
Before we construct the actual CARE model, we have to define its configuration via a Config
object, which includes
The defaults should be sensible in many cases, so a change should only be necessary if the training process fails.
Important: Note that for this notebook we use a very small number of update steps per epoch for immediate feedback, whereas this number should be increased considerably (e.g. train_steps_per_epoch=400
, train_batch_size=16
) to obtain a well-trained model.
config = Config(axes, n_channel_in, n_channel_out, train_steps_per_epoch=25, train_batch_size=4)
print(config)
vars(config)
Config(axes='ZYXC', n_channel_in=1, n_channel_out=1, n_dim=3, probabilistic=False, train_batch_size=4, train_checkpoint='weights_best.h5', train_checkpoint_epoch='weights_now.h5', train_checkpoint_last='weights_last.h5', train_epochs=100, train_learning_rate=0.0004, train_loss='mae', train_reduce_lr={'factor': 0.5, 'patience': 10, 'min_delta': 0}, train_steps_per_epoch=25, train_tensorboard=True, unet_input_shape=(None, None, None, 1), unet_kern_size=3, unet_last_activation='linear', unet_n_depth=2, unet_n_first=32, unet_residual=True)
{'n_dim': 3, 'axes': 'ZYXC', 'n_channel_in': 1, 'n_channel_out': 1, 'train_checkpoint': 'weights_best.h5', 'train_checkpoint_last': 'weights_last.h5', 'train_checkpoint_epoch': 'weights_now.h5', 'probabilistic': False, 'unet_residual': True, 'unet_n_depth': 2, 'unet_kern_size': 3, 'unet_n_first': 32, 'unet_last_activation': 'linear', 'unet_input_shape': (None, None, None, 1), 'train_loss': 'mae', 'train_epochs': 100, 'train_steps_per_epoch': 25, 'train_learning_rate': 0.0004, 'train_batch_size': 4, 'train_tensorboard': True, 'train_reduce_lr': {'factor': 0.5, 'patience': 10, 'min_delta': 0}}
We now create an upsampling CARE model with the chosen configuration:
model = UpsamplingCARE(config, 'my_model', basedir='models')
Training the model will likely take some time. We recommend to monitor the progress with TensorBoard (example below), which allows you to inspect the losses during training. Furthermore, you can look at the predictions for some of the validation images, which can be helpful to recognize problems early on.
You can start TensorBoard from the current working directory with tensorboard --logdir=.
Then connect to http://localhost:6006/ with your browser.
history = model.train(X,Y, validation_data=(X_val,Y_val))
Epoch 1/100 WARNING:tensorflow:AutoGraph could not transform <function _mean_or_not.<locals>.<lambda> at 0x7f790788d430> and will run it as-is. Cause: could not parse the source code of <function _mean_or_not.<locals>.<lambda> at 0x7f790788d430>: found multiple definitions with identical signatures at the location. This error may be avoided by defining each lambda on a single line and with unique argument names. The matching definitions were: Match 0: (lambda x: K.mean(x, axis=(- 1))) Match 1: (lambda x: x) To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert WARNING: AutoGraph could not transform <function _mean_or_not.<locals>.<lambda> at 0x7f790788d430> and will run it as-is. Cause: could not parse the source code of <function _mean_or_not.<locals>.<lambda> at 0x7f790788d430>: found multiple definitions with identical signatures at the location. This error may be avoided by defining each lambda on a single line and with unique argument names. The matching definitions were: Match 0: (lambda x: K.mean(x, axis=(- 1))) Match 1: (lambda x: x) To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert WARNING:tensorflow:AutoGraph could not transform <function _mean_or_not.<locals>.<lambda> at 0x7f790788d670> and will run it as-is. Cause: could not parse the source code of <function _mean_or_not.<locals>.<lambda> at 0x7f790788d670>: found multiple definitions with identical signatures at the location. This error may be avoided by defining each lambda on a single line and with unique argument names. The matching definitions were: Match 0: (lambda x: K.mean(x, axis=(- 1))) Match 1: (lambda x: x) To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert WARNING: AutoGraph could not transform <function _mean_or_not.<locals>.<lambda> at 0x7f790788d670> and will run it as-is. Cause: could not parse the source code of <function _mean_or_not.<locals>.<lambda> at 0x7f790788d670>: found multiple definitions with identical signatures at the location. This error may be avoided by defining each lambda on a single line and with unique argument names. The matching definitions were: Match 0: (lambda x: K.mean(x, axis=(- 1))) Match 1: (lambda x: x) To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert WARNING:tensorflow:AutoGraph could not transform <function _mean_or_not.<locals>.<lambda> at 0x7f790788d790> and will run it as-is. Cause: could not parse the source code of <function _mean_or_not.<locals>.<lambda> at 0x7f790788d790>: found multiple definitions with identical signatures at the location. This error may be avoided by defining each lambda on a single line and with unique argument names. The matching definitions were: Match 0: (lambda x: K.mean(x, axis=(- 1))) Match 1: (lambda x: x) To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert WARNING: AutoGraph could not transform <function _mean_or_not.<locals>.<lambda> at 0x7f790788d790> and will run it as-is. Cause: could not parse the source code of <function _mean_or_not.<locals>.<lambda> at 0x7f790788d790>: found multiple definitions with identical signatures at the location. This error may be avoided by defining each lambda on a single line and with unique argument names. The matching definitions were: Match 0: (lambda x: K.mean(x, axis=(- 1))) Match 1: (lambda x: x) To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert 6/25 [======>.......................] - ETA: 2s - loss: 0.1334 - mse: 0.0281 - mae: 0.1334WARNING:tensorflow:Callback method `on_train_batch_end` is slow compared to the batch time (batch time: 0.0398s vs `on_train_batch_end` time: 0.0747s). Check your callbacks. 25/25 [==============================] - 11s 349ms/step - loss: 0.1166 - mse: 0.0222 - mae: 0.1166 - val_loss: 0.1039 - val_mse: 0.0187 - val_mae: 0.1039 - lr: 4.0000e-04 Epoch 2/100 25/25 [==============================] - 5s 200ms/step - loss: 0.0914 - mse: 0.0145 - mae: 0.0914 - val_loss: 0.0772 - val_mse: 0.0104 - val_mae: 0.0772 - lr: 4.0000e-04 Epoch 3/100 25/25 [==============================] - 5s 200ms/step - loss: 0.0706 - mse: 0.0091 - mae: 0.0706 - val_loss: 0.0680 - val_mse: 0.0087 - val_mae: 0.0680 - lr: 4.0000e-04 Epoch 4/100 25/25 [==============================] - 5s 200ms/step - loss: 0.0679 - mse: 0.0087 - mae: 0.0679 - val_loss: 0.0652 - val_mse: 0.0080 - val_mae: 0.0652 - lr: 4.0000e-04 Epoch 5/100 25/25 [==============================] - 5s 204ms/step - loss: 0.0623 - mse: 0.0077 - mae: 0.0623 - val_loss: 0.0594 - val_mse: 0.0072 - val_mae: 0.0594 - lr: 4.0000e-04 Epoch 6/100 25/25 [==============================] - 5s 199ms/step - loss: 0.0592 - mse: 0.0073 - mae: 0.0592 - val_loss: 0.0595 - val_mse: 0.0078 - val_mae: 0.0595 - lr: 4.0000e-04 Epoch 7/100 25/25 [==============================] - 5s 200ms/step - loss: 0.0581 - mse: 0.0071 - mae: 0.0581 - val_loss: 0.0583 - val_mse: 0.0077 - val_mae: 0.0583 - lr: 4.0000e-04 Epoch 8/100 25/25 [==============================] - 5s 200ms/step - loss: 0.0532 - mse: 0.0064 - mae: 0.0532 - val_loss: 0.0505 - val_mse: 0.0061 - val_mae: 0.0505 - lr: 4.0000e-04 Epoch 9/100 25/25 [==============================] - 5s 200ms/step - loss: 0.0489 - mse: 0.0056 - mae: 0.0489 - val_loss: 0.0495 - val_mse: 0.0056 - val_mae: 0.0495 - lr: 4.0000e-04 Epoch 10/100 25/25 [==============================] - 5s 200ms/step - loss: 0.0522 - mse: 0.0058 - mae: 0.0522 - val_loss: 0.0571 - val_mse: 0.0062 - val_mae: 0.0571 - lr: 4.0000e-04 Epoch 11/100 25/25 [==============================] - 5s 200ms/step - loss: 0.0491 - mse: 0.0054 - mae: 0.0491 - val_loss: 0.0445 - val_mse: 0.0048 - val_mae: 0.0445 - lr: 4.0000e-04 Epoch 12/100 25/25 [==============================] - 5s 200ms/step - loss: 0.0441 - mse: 0.0046 - mae: 0.0441 - val_loss: 0.0408 - val_mse: 0.0040 - val_mae: 0.0408 - lr: 4.0000e-04 Epoch 13/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0400 - mse: 0.0038 - mae: 0.0400 - val_loss: 0.0393 - val_mse: 0.0037 - val_mae: 0.0393 - lr: 4.0000e-04 Epoch 14/100 25/25 [==============================] - 5s 200ms/step - loss: 0.0381 - mse: 0.0035 - mae: 0.0381 - val_loss: 0.0377 - val_mse: 0.0035 - val_mae: 0.0377 - lr: 4.0000e-04 Epoch 15/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0367 - mse: 0.0033 - mae: 0.0367 - val_loss: 0.0360 - val_mse: 0.0033 - val_mae: 0.0360 - lr: 4.0000e-04 Epoch 16/100 25/25 [==============================] - 5s 200ms/step - loss: 0.0366 - mse: 0.0032 - mae: 0.0366 - val_loss: 0.0376 - val_mse: 0.0034 - val_mae: 0.0376 - lr: 4.0000e-04 Epoch 17/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0355 - mse: 0.0031 - mae: 0.0355 - val_loss: 0.0342 - val_mse: 0.0030 - val_mae: 0.0342 - lr: 4.0000e-04 Epoch 18/100 25/25 [==============================] - 5s 200ms/step - loss: 0.0345 - mse: 0.0030 - mae: 0.0345 - val_loss: 0.0336 - val_mse: 0.0029 - val_mae: 0.0336 - lr: 4.0000e-04 Epoch 19/100 25/25 [==============================] - 5s 199ms/step - loss: 0.0345 - mse: 0.0029 - mae: 0.0345 - val_loss: 0.0357 - val_mse: 0.0030 - val_mae: 0.0357 - lr: 4.0000e-04 Epoch 20/100 25/25 [==============================] - 5s 200ms/step - loss: 0.0350 - mse: 0.0029 - mae: 0.0350 - val_loss: 0.0390 - val_mse: 0.0031 - val_mae: 0.0390 - lr: 4.0000e-04 Epoch 21/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0340 - mse: 0.0028 - mae: 0.0340 - val_loss: 0.0321 - val_mse: 0.0027 - val_mae: 0.0321 - lr: 4.0000e-04 Epoch 22/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0318 - mse: 0.0025 - mae: 0.0318 - val_loss: 0.0336 - val_mse: 0.0027 - val_mae: 0.0336 - lr: 4.0000e-04 Epoch 23/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0321 - mse: 0.0026 - mae: 0.0321 - val_loss: 0.0317 - val_mse: 0.0025 - val_mae: 0.0317 - lr: 4.0000e-04 Epoch 24/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0321 - mse: 0.0026 - mae: 0.0321 - val_loss: 0.0306 - val_mse: 0.0025 - val_mae: 0.0306 - lr: 4.0000e-04 Epoch 25/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0324 - mse: 0.0027 - mae: 0.0324 - val_loss: 0.0331 - val_mse: 0.0026 - val_mae: 0.0331 - lr: 4.0000e-04 Epoch 26/100 25/25 [==============================] - 5s 200ms/step - loss: 0.0307 - mse: 0.0024 - mae: 0.0307 - val_loss: 0.0307 - val_mse: 0.0025 - val_mae: 0.0307 - lr: 4.0000e-04 Epoch 27/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0298 - mse: 0.0023 - mae: 0.0298 - val_loss: 0.0308 - val_mse: 0.0026 - val_mae: 0.0308 - lr: 4.0000e-04 Epoch 28/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0297 - mse: 0.0023 - mae: 0.0297 - val_loss: 0.0295 - val_mse: 0.0023 - val_mae: 0.0295 - lr: 4.0000e-04 Epoch 29/100 25/25 [==============================] - 5s 200ms/step - loss: 0.0287 - mse: 0.0022 - mae: 0.0287 - val_loss: 0.0299 - val_mse: 0.0023 - val_mae: 0.0299 - lr: 4.0000e-04 Epoch 30/100 25/25 [==============================] - 5s 200ms/step - loss: 0.0308 - mse: 0.0023 - mae: 0.0308 - val_loss: 0.0317 - val_mse: 0.0026 - val_mae: 0.0317 - lr: 4.0000e-04 Epoch 31/100 25/25 [==============================] - 5s 202ms/step - loss: 0.0297 - mse: 0.0022 - mae: 0.0297 - val_loss: 0.0286 - val_mse: 0.0022 - val_mae: 0.0286 - lr: 4.0000e-04 Epoch 32/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0303 - mse: 0.0024 - mae: 0.0303 - val_loss: 0.0300 - val_mse: 0.0022 - val_mae: 0.0300 - lr: 4.0000e-04 Epoch 33/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0288 - mse: 0.0021 - mae: 0.0288 - val_loss: 0.0279 - val_mse: 0.0021 - val_mae: 0.0279 - lr: 4.0000e-04 Epoch 34/100 25/25 [==============================] - 5s 200ms/step - loss: 0.0302 - mse: 0.0023 - mae: 0.0302 - val_loss: 0.0304 - val_mse: 0.0022 - val_mae: 0.0304 - lr: 4.0000e-04 Epoch 35/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0288 - mse: 0.0021 - mae: 0.0288 - val_loss: 0.0281 - val_mse: 0.0021 - val_mae: 0.0281 - lr: 4.0000e-04 Epoch 36/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0276 - mse: 0.0021 - mae: 0.0276 - val_loss: 0.0286 - val_mse: 0.0022 - val_mae: 0.0286 - lr: 4.0000e-04 Epoch 37/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0286 - mse: 0.0022 - mae: 0.0286 - val_loss: 0.0296 - val_mse: 0.0021 - val_mae: 0.0296 - lr: 4.0000e-04 Epoch 38/100 25/25 [==============================] - 5s 202ms/step - loss: 0.0280 - mse: 0.0020 - mae: 0.0280 - val_loss: 0.0276 - val_mse: 0.0021 - val_mae: 0.0276 - lr: 4.0000e-04 Epoch 39/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0270 - mse: 0.0019 - mae: 0.0270 - val_loss: 0.0274 - val_mse: 0.0021 - val_mae: 0.0274 - lr: 4.0000e-04 Epoch 40/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0277 - mse: 0.0020 - mae: 0.0277 - val_loss: 0.0273 - val_mse: 0.0020 - val_mae: 0.0273 - lr: 4.0000e-04 Epoch 41/100 25/25 [==============================] - 5s 202ms/step - loss: 0.0266 - mse: 0.0019 - mae: 0.0266 - val_loss: 0.0265 - val_mse: 0.0019 - val_mae: 0.0265 - lr: 4.0000e-04 Epoch 42/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0273 - mse: 0.0019 - mae: 0.0273 - val_loss: 0.0270 - val_mse: 0.0019 - val_mae: 0.0270 - lr: 4.0000e-04 Epoch 43/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0259 - mse: 0.0018 - mae: 0.0259 - val_loss: 0.0277 - val_mse: 0.0021 - val_mae: 0.0277 - lr: 4.0000e-04 Epoch 44/100 25/25 [==============================] - 5s 202ms/step - loss: 0.0261 - mse: 0.0019 - mae: 0.0261 - val_loss: 0.0264 - val_mse: 0.0019 - val_mae: 0.0264 - lr: 4.0000e-04 Epoch 45/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0263 - mse: 0.0019 - mae: 0.0263 - val_loss: 0.0269 - val_mse: 0.0019 - val_mae: 0.0269 - lr: 4.0000e-04 Epoch 46/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0274 - mse: 0.0019 - mae: 0.0274 - val_loss: 0.0271 - val_mse: 0.0020 - val_mae: 0.0271 - lr: 4.0000e-04 Epoch 47/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0275 - mse: 0.0019 - mae: 0.0275 - val_loss: 0.0261 - val_mse: 0.0019 - val_mae: 0.0261 - lr: 4.0000e-04 Epoch 48/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0260 - mse: 0.0018 - mae: 0.0260 - val_loss: 0.0267 - val_mse: 0.0019 - val_mae: 0.0267 - lr: 4.0000e-04 Epoch 49/100 25/25 [==============================] - 5s 202ms/step - loss: 0.0254 - mse: 0.0018 - mae: 0.0254 - val_loss: 0.0254 - val_mse: 0.0018 - val_mae: 0.0254 - lr: 4.0000e-04 Epoch 50/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0256 - mse: 0.0018 - mae: 0.0256 - val_loss: 0.0276 - val_mse: 0.0020 - val_mae: 0.0276 - lr: 4.0000e-04 Epoch 51/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0254 - mse: 0.0017 - mae: 0.0254 - val_loss: 0.0257 - val_mse: 0.0018 - val_mae: 0.0257 - lr: 4.0000e-04 Epoch 52/100 25/25 [==============================] - 5s 202ms/step - loss: 0.0259 - mse: 0.0018 - mae: 0.0259 - val_loss: 0.0254 - val_mse: 0.0018 - val_mae: 0.0254 - lr: 4.0000e-04 Epoch 53/100 25/25 [==============================] - 5s 202ms/step - loss: 0.0259 - mse: 0.0018 - mae: 0.0259 - val_loss: 0.0250 - val_mse: 0.0018 - val_mae: 0.0250 - lr: 4.0000e-04 Epoch 54/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0253 - mse: 0.0017 - mae: 0.0253 - val_loss: 0.0244 - val_mse: 0.0017 - val_mae: 0.0244 - lr: 4.0000e-04 Epoch 55/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0271 - mse: 0.0018 - mae: 0.0271 - val_loss: 0.0276 - val_mse: 0.0019 - val_mae: 0.0276 - lr: 4.0000e-04 Epoch 56/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0266 - mse: 0.0018 - mae: 0.0266 - val_loss: 0.0250 - val_mse: 0.0017 - val_mae: 0.0250 - lr: 4.0000e-04 Epoch 57/100 25/25 [==============================] - 5s 200ms/step - loss: 0.0245 - mse: 0.0016 - mae: 0.0245 - val_loss: 0.0248 - val_mse: 0.0017 - val_mae: 0.0248 - lr: 4.0000e-04 Epoch 58/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0249 - mse: 0.0017 - mae: 0.0249 - val_loss: 0.0249 - val_mse: 0.0017 - val_mae: 0.0249 - lr: 4.0000e-04 Epoch 59/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0247 - mse: 0.0017 - mae: 0.0247 - val_loss: 0.0246 - val_mse: 0.0017 - val_mae: 0.0246 - lr: 4.0000e-04 Epoch 60/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0259 - mse: 0.0017 - mae: 0.0259 - val_loss: 0.0266 - val_mse: 0.0020 - val_mae: 0.0266 - lr: 4.0000e-04 Epoch 61/100 25/25 [==============================] - 5s 202ms/step - loss: 0.0238 - mse: 0.0016 - mae: 0.0238 - val_loss: 0.0244 - val_mse: 0.0017 - val_mae: 0.0244 - lr: 4.0000e-04 Epoch 62/100 25/25 [==============================] - 5s 202ms/step - loss: 0.0235 - mse: 0.0015 - mae: 0.0235 - val_loss: 0.0236 - val_mse: 0.0016 - val_mae: 0.0236 - lr: 4.0000e-04 Epoch 63/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0233 - mse: 0.0015 - mae: 0.0233 - val_loss: 0.0241 - val_mse: 0.0017 - val_mae: 0.0241 - lr: 4.0000e-04 Epoch 64/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0248 - mse: 0.0016 - mae: 0.0248 - val_loss: 0.0267 - val_mse: 0.0017 - val_mae: 0.0267 - lr: 4.0000e-04 Epoch 65/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0244 - mse: 0.0016 - mae: 0.0244 - val_loss: 0.0248 - val_mse: 0.0016 - val_mae: 0.0248 - lr: 4.0000e-04 Epoch 66/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0238 - mse: 0.0016 - mae: 0.0238 - val_loss: 0.0237 - val_mse: 0.0016 - val_mae: 0.0237 - lr: 4.0000e-04 Epoch 67/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0256 - mse: 0.0017 - mae: 0.0256 - val_loss: 0.0265 - val_mse: 0.0017 - val_mae: 0.0265 - lr: 4.0000e-04 Epoch 68/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0244 - mse: 0.0016 - mae: 0.0244 - val_loss: 0.0236 - val_mse: 0.0016 - val_mae: 0.0236 - lr: 4.0000e-04 Epoch 69/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0238 - mse: 0.0015 - mae: 0.0238 - val_loss: 0.0237 - val_mse: 0.0016 - val_mae: 0.0237 - lr: 4.0000e-04 Epoch 70/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0224 - mse: 0.0014 - mae: 0.0224 - val_loss: 0.0233 - val_mse: 0.0015 - val_mae: 0.0233 - lr: 4.0000e-04 Epoch 71/100 25/25 [==============================] - 5s 202ms/step - loss: 0.0225 - mse: 0.0014 - mae: 0.0225 - val_loss: 0.0228 - val_mse: 0.0015 - val_mae: 0.0228 - lr: 4.0000e-04 Epoch 72/100 25/25 [==============================] - 5s 202ms/step - loss: 0.0221 - mse: 0.0014 - mae: 0.0221 - val_loss: 0.0225 - val_mse: 0.0015 - val_mae: 0.0225 - lr: 4.0000e-04 Epoch 73/100 25/25 [==============================] - 5s 202ms/step - loss: 0.0219 - mse: 0.0013 - mae: 0.0219 - val_loss: 0.0224 - val_mse: 0.0015 - val_mae: 0.0224 - lr: 4.0000e-04 Epoch 74/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0220 - mse: 0.0014 - mae: 0.0220 - val_loss: 0.0223 - val_mse: 0.0014 - val_mae: 0.0223 - lr: 4.0000e-04 Epoch 75/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0224 - mse: 0.0014 - mae: 0.0224 - val_loss: 0.0218 - val_mse: 0.0014 - val_mae: 0.0218 - lr: 4.0000e-04 Epoch 76/100 25/25 [==============================] - 5s 200ms/step - loss: 0.0219 - mse: 0.0014 - mae: 0.0219 - val_loss: 0.0223 - val_mse: 0.0014 - val_mae: 0.0223 - lr: 4.0000e-04 Epoch 77/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0224 - mse: 0.0014 - mae: 0.0224 - val_loss: 0.0229 - val_mse: 0.0015 - val_mae: 0.0229 - lr: 4.0000e-04 Epoch 78/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0222 - mse: 0.0014 - mae: 0.0222 - val_loss: 0.0216 - val_mse: 0.0014 - val_mae: 0.0216 - lr: 4.0000e-04 Epoch 79/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0224 - mse: 0.0014 - mae: 0.0224 - val_loss: 0.0222 - val_mse: 0.0014 - val_mae: 0.0222 - lr: 4.0000e-04 Epoch 80/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0218 - mse: 0.0013 - mae: 0.0218 - val_loss: 0.0217 - val_mse: 0.0014 - val_mae: 0.0217 - lr: 4.0000e-04 Epoch 81/100 25/25 [==============================] - 5s 200ms/step - loss: 0.0226 - mse: 0.0014 - mae: 0.0226 - val_loss: 0.0223 - val_mse: 0.0014 - val_mae: 0.0223 - lr: 4.0000e-04 Epoch 82/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0215 - mse: 0.0013 - mae: 0.0215 - val_loss: 0.0214 - val_mse: 0.0014 - val_mae: 0.0214 - lr: 4.0000e-04 Epoch 83/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0212 - mse: 0.0013 - mae: 0.0212 - val_loss: 0.0211 - val_mse: 0.0013 - val_mae: 0.0211 - lr: 4.0000e-04 Epoch 84/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0204 - mse: 0.0012 - mae: 0.0204 - val_loss: 0.0216 - val_mse: 0.0013 - val_mae: 0.0216 - lr: 4.0000e-04 Epoch 85/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0217 - mse: 0.0013 - mae: 0.0217 - val_loss: 0.0224 - val_mse: 0.0014 - val_mae: 0.0224 - lr: 4.0000e-04 Epoch 86/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0220 - mse: 0.0013 - mae: 0.0220 - val_loss: 0.0240 - val_mse: 0.0014 - val_mae: 0.0240 - lr: 4.0000e-04 Epoch 87/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0219 - mse: 0.0013 - mae: 0.0219 - val_loss: 0.0212 - val_mse: 0.0013 - val_mae: 0.0212 - lr: 4.0000e-04 Epoch 88/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0204 - mse: 0.0012 - mae: 0.0204 - val_loss: 0.0208 - val_mse: 0.0013 - val_mae: 0.0208 - lr: 4.0000e-04 Epoch 89/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0224 - mse: 0.0013 - mae: 0.0224 - val_loss: 0.0231 - val_mse: 0.0014 - val_mae: 0.0231 - lr: 4.0000e-04 Epoch 90/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0214 - mse: 0.0012 - mae: 0.0214 - val_loss: 0.0209 - val_mse: 0.0013 - val_mae: 0.0209 - lr: 4.0000e-04 Epoch 91/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0205 - mse: 0.0012 - mae: 0.0205 - val_loss: 0.0205 - val_mse: 0.0013 - val_mae: 0.0205 - lr: 4.0000e-04 Epoch 92/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0209 - mse: 0.0012 - mae: 0.0209 - val_loss: 0.0243 - val_mse: 0.0015 - val_mae: 0.0243 - lr: 4.0000e-04 Epoch 93/100 25/25 [==============================] - 5s 200ms/step - loss: 0.0239 - mse: 0.0014 - mae: 0.0239 - val_loss: 0.0251 - val_mse: 0.0015 - val_mae: 0.0251 - lr: 4.0000e-04 Epoch 94/100 25/25 [==============================] - 5s 200ms/step - loss: 0.0221 - mse: 0.0013 - mae: 0.0221 - val_loss: 0.0205 - val_mse: 0.0013 - val_mae: 0.0205 - lr: 4.0000e-04 Epoch 95/100 25/25 [==============================] - 5s 200ms/step - loss: 0.0204 - mse: 0.0012 - mae: 0.0204 - val_loss: 0.0209 - val_mse: 0.0013 - val_mae: 0.0209 - lr: 4.0000e-04 Epoch 96/100 25/25 [==============================] - 5s 200ms/step - loss: 0.0212 - mse: 0.0012 - mae: 0.0212 - val_loss: 0.0233 - val_mse: 0.0014 - val_mae: 0.0233 - lr: 4.0000e-04 Epoch 97/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0210 - mse: 0.0012 - mae: 0.0210 - val_loss: 0.0202 - val_mse: 0.0012 - val_mae: 0.0202 - lr: 4.0000e-04 Epoch 98/100 25/25 [==============================] - 5s 200ms/step - loss: 0.0196 - mse: 0.0011 - mae: 0.0196 - val_loss: 0.0213 - val_mse: 0.0013 - val_mae: 0.0213 - lr: 4.0000e-04 Epoch 99/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0202 - mse: 0.0011 - mae: 0.0202 - val_loss: 0.0227 - val_mse: 0.0013 - val_mae: 0.0227 - lr: 4.0000e-04 Epoch 100/100 25/25 [==============================] - 5s 201ms/step - loss: 0.0208 - mse: 0.0012 - mae: 0.0208 - val_loss: 0.0213 - val_mse: 0.0013 - val_mae: 0.0213 - lr: 4.0000e-04 Loading network weights from 'weights_best.h5'.
Plot final training history (available in TensorBoard during training):
print(sorted(list(history.history.keys())))
plt.figure(figsize=(16,5))
plot_history(history,['loss','val_loss'],['mse','val_mse','mae','val_mae']);
['loss', 'lr', 'mae', 'mse', 'val_loss', 'val_mae', 'val_mse']
plt.figure(figsize=(12,4.5))
_P = model.keras_model.predict(X_val[:5])
if config.probabilistic:
_P = _P[...,:(_P.shape[-1]//2)]
plot_some(X_val[:5,...,0,0],Y_val[:5,...,0,0],_P[...,0,0],pmax=99.5)
plt.suptitle('5 example validation patches (ZY slice)\n'
'top row: input (source), '
'middle row: target (ground truth), '
'bottom row: predicted from source');
See https://github.com/CSBDeep/CSBDeep_website/wiki/Your-Model-in-Fiji for details.
model.export_TF()
WARNING:tensorflow:From /home/uwe/sw/miniconda3/envs/ws/lib/python3.8/site-packages/tensorflow/python/saved_model/signature_def_utils_impl.py:203: build_tensor_info (from tensorflow.python.saved_model.utils_impl) is deprecated and will be removed in a future version. Instructions for updating: This function will only be available through the v1 compatibility library as tf.compat.v1.saved_model.utils.build_tensor_info or tf.compat.v1.saved_model.build_tensor_info. INFO:tensorflow:No assets to save. INFO:tensorflow:No assets to write. INFO:tensorflow:SavedModel written to: /tmp/tmp878siqwt/model/saved_model.pb Model exported in TensorFlow's SavedModel format: /home/uwe/research/csbdeep/examples/examples/upsampling3D/models/my_model/TF_SavedModel.zip
***IMPORTANT NOTE*** You are using 'tensorflow' 2.x, hence it is likely that the exported model *will not work* in associated ImageJ/Fiji plugins (e.g. CSBDeep and StarDist). If you indeed have problems loading the exported model in Fiji, the current workaround is to load the trained model in a Python environment with installed 'tensorflow' 1.x and then export it again. If you need help with this, please read: https://gist.github.com/uschmidt83/4b747862fe307044c722d6d1009f6183