Tutorial Overview¶
On completion of this tutorial you will have learnt how to write your own blind-spot denoising procedure that is trained in a self-supervised manner, i.e., the training data is the same as the inference data with no labels required!
Methodology Recap¶
We will implement the Noise2Void methodology of blind-spot networks for denoising. This involves performing a pre-processing step which identifies the 'active' pixels and then replaces their original noisy value with that of a neighbouring pixel. This processed data becomes the input to the neural network with the original noisy image being the network's target. However, unlike in most NN applications, the loss is not computed across the full predicted image, but only at the active pixel locations.
# Import necessary basic packages
import numpy as np
%matplotlib inline
import matplotlib.pyplot as plt
from tqdm import tqdm
# Import necessary torch packages
import torch
import torch.nn as nn
from torch.utils.data import TensorDataset, DataLoader
# Import our pre-made functions which will keep the notebook concise
# These functions are independent to the blindspot application however important for the data handling and
# network creation/initialisation
from unet import UNet
from tutorial_utils import regular_patching_2D, add_whitegaussian_noise, weights_init, set_seed, make_data_loader, plot_corruption, plot_training_metrics, plot_synth_results
# Some general plotting parameters so we don't need to keep adding them later on
cmap='RdBu'
vmin = -0.25
vmax = 0.25
# For reproducibility purposes we set random, numpy and torch seeds
set_seed(42) TrueStep One - Data loading¶
In this example we are going to use a post-stack seismic section generated from the Hess VTI model. The post-stack section is available in the public data folder: https://exrcsdrive.kaust.edu.sa/exrcsdrive/index.php/s/vjjry6BZ3n3Ewei
with password: kaust
If the folder is no longer public, this is likely due to expired rights. Please email: cebirnie[at]kaust.edu.sa to request access.
In this instance I have downloaded the file and added to a folder in this repository title 'data'.
d = np.load("../data/Hess_PostStackSection.npy")
# Check data dimensions
print(d.shape)(196, 452)
Plot the data to see what it looks like¶
plt.figure(figsize=[7,5])
plt.imshow(d, cmap=cmap, vmin=vmin, vmax=vmax)<matplotlib.image.AxesImage at 0x2afaac342650>
Add noise¶
As we can see from above, the data which you loaded in is the noise-free synthetic. This is great for helping us benchmark the results however we are really interested in testing the denoising performance of blind-spot networks there we need to add some noise that we wish to later suppress.
noisydata, _ = add_whitegaussian_noise(d, sc=0.1)Plot the noisy data to see what it looks like in comparison to the clean¶
plt.figure(figsize=[7,5])
plt.imshow(noisydata, cmap=cmap, vmin=vmin, vmax=vmax)<matplotlib.image.AxesImage at 0x2afaac4646d0>
Patch data¶
At the moment we have a single image that we wish to denoise however to train the network we need to give it multiple data examples. Therefore, following common computer vision methodology, we will select random patches from the data for the networks training.
Our patch implementation involves first regularly extracting patches from the image and then shuffling the patches such that they are in a random order. Later at the training stage these patches will be split into train and test dataset.
# Regularly extract patches from the noisy data
noisy_patches = regular_patching_2D(noisydata,
patchsize=[32, 32], # dimensions of extracted patch
step=[4,6], # step to be taken in y,x for the extraction procedure
)
# Randomise patch order
shuffler = np.random.permutation(len(noisy_patches))
noisy_patches = noisy_patches[shuffler] Extracting 2870 patches
Visualise the training patches¶
fig, axs = plt.subplots(3,6,figsize=[15,7])
for i in range(6*3):
axs.ravel()[i].imshow(noisy_patches[i], cmap=cmap, vmin=vmin, vmax=vmax)
fig.tight_layout()
Step Two - Blindspot corruption of training data¶
Now we have made our noisy data into patches such that we have an adequate number to train the network, we now need to pre-process these noisy patches prior to being input into the network.
Our implementation of the preprocessing involves:
- selecting the active pixels
- selecting the neighbourhood pixel for each active pixel, which it will take the value of
- replacing each active pixels' value with its neighbourhood pixels' value
- creating a active pixel 'mask' which shows the location of the active pixels on the patchThe first three steps are important for the pre-processing of the noisy patches, whilst the fourth step is required for identifying the locations on which to compute the loss function during training.
To do: Create a function that randomly selects and corrupts pixels following N2V methodology¶
def multi_active_pixels(patch,
num_activepixels,
neighbourhood_radius=5,
):
""" Function to identify multiple active pixels and replace with values from neighbouring pixels
Parameters
----------
patch : numpy 2D array
Noisy patch of data to be processed
num_activepixels : int
Number of active pixels to be selected within the patch
neighbourhood_radius : int
Radius over which to select neighbouring pixels for active pixel value replacement
Returns
-------
cp_ptch : numpy 2D array
Processed patch
mask : numpy 2D array
Mask showing location of active pixels within the patch
"""
n_rad = neighbourhood_radius # descriptive variable name was a little long
# ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
# STEP ONE: SELECT ACTIVE PIXEL LOCATIONS
idx_aps = np.random.randint(0, patch.shape[0], num_activepixels)
idy_aps = np.random.randint(0, patch.shape[1], num_activepixels)
id_aps = (idx_aps, idy_aps)
# ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
# STEP TWO: SELECT NEIGHBOURING PIXEL LOCATIONS
# PART 1: Compute Shift
# For each active pixel compute shift for finding neighbouring pixel and find pixel
x_neigh_shft = np.random.randint(-n_rad // 2 + n_rad % 2, n_rad // 2 + n_rad % 2, num_activepixels)
y_neigh_shft = np.random.randint(-n_rad // 2 + n_rad % 2, n_rad // 2 + n_rad % 2, num_activepixels)
# OPTIONAL: don't allow replacement with itself
for i in range(len(x_neigh_shft)):
if x_neigh_shft[i] == 0 and y_neigh_shft[i] == 0:
# This means its replacing itself with itself...
shft_options = np.trim_zeros(np.arange(-n_rad // 2 + 1, n_rad // 2 + 1))
x_neigh_shft[i] = np.random.choice(shft_options[shft_options != 0], 1)
# PART 2: Find x and y locations of neighbours for the replacement
idx_neigh = idx_aps + x_neigh_shft
idy_neigh = idy_aps + y_neigh_shft
# Ensure neighbouring pixels within patch window
idx_neigh = idx_neigh + (idx_neigh < 0) * patch.shape[0] - (idx_neigh >= patch.shape[0]) * patch.shape[0]
idy_neigh = idy_neigh + (idy_neigh < 0) * patch.shape[1] - (idy_neigh >= patch.shape[1]) * patch.shape[1]
# Get x,y of neighbouring pixels
id_neigh = (idx_neigh, idy_neigh)
# ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
# STEP THREE: REPLACE ACTIVE PIXEL VALUES BY NEIGHBOURS
cp_ptch = patch.copy()
cp_ptch[id_aps] = patch[id_neigh]
# ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
# STEP FOUR: MAKE ACTIVE PIXEL MASK
mask = np.ones_like(patch)
mask[id_aps] = 0.
return cp_ptch, mask
TO DO: CHECK THE CORRUPTION FUNCTION WORKS¶
# Input the values of your choice into your pre-processing function
crpt_patch, mask = multi_active_pixels(noisy_patches[6],
num_activepixels=50,
neighbourhood_radius=5,
)
# Use the pre-made plotting function to visualise the corruption
fig,axs = plot_corruption(noisy_patches[6], crpt_patch, mask)
TO DO: SELECT THE NUMBER OF ACTIVE PIXELS (AS PERCENTAGE)¶
In the original N2V examples the authors use between 0.5 and 2% for the number of active pixels within a patch.
In Birnie et al., 2021 where they use this methodology for the suppression of white, Gaussian noise, the authors use 0.2%. However, in their example they have substantially more training patches.
# Choose the percent of active pixels per patch
perc_active = 2
# Compute the total number of pixels within a patch
total_num_pixels = noisy_patches[0].shape[0]*noisy_patches[0].shape[1]
# Compute the number that should be active pixels based on the choosen percentage
num_activepixels = int(np.floor((total_num_pixels/100) * perc_active))
print("Number of active pixels selected: \n %.2f percent equals %i pixels"%(perc_active,num_activepixels))
# Input the values of your choice into your pre-processing function
crpt_patch, mask = multi_active_pixels(noisy_patches[6],
num_activepixels=num_activepixels,
neighbourhood_radius=5,
)
# Visulise the coverage of active pixels within a patch
fig,axs = plot_corruption(noisy_patches[6], crpt_patch, mask)Number of active pixels selected:
2.00 percent equals 20 pixels
Step three - Set up network¶
In the N2V application of Krull et al., 2018, the network is not specially tailored to the blindspot task. As such, in theory any network could be used that goes from one input image to another of the same size.
In this example, like in Krull et al., 2018 and Birnie et al., 2021's seismic application, we will use a standard UNet architecture. As the architecture is independent to the blind-spot denoising procedure presented, it will be created via functions as opposed to being wrote within the notebook.
# Select device for training
device = 'cpu'
if torch.cuda.device_count() > 0 and torch.cuda.is_available():
print("Cuda installed! Running on GPU!")
device = torch.device(torch.cuda.current_device())
print(f'Device: {device} {torch.cuda.get_device_name(device)}')
else:
print("No GPU available!")Cuda installed! Running on GPU!
Device: cuda:0 NVIDIA Tesla V100-SXM2-32GB
Build the network¶
# Build UNet from pre-made function
network = UNet(input_channels=1,
output_channels=1,
hidden_channels=32,
levels=2).to(device)
# Initialise UNet's weights from pre-made function
network = network.apply(weights_init) /ibex/scratch/birniece/Transform2022_SelfSupervisedDenoising/Solutions/tutorial_utils.py:267: UserWarning: nn.init.xavier_normal is now deprecated in favor of nn.init.xavier_normal_.
nn.init.xavier_normal(m.weight)
/ibex/scratch/birniece/Transform2022_SelfSupervisedDenoising/Solutions/tutorial_utils.py:268: UserWarning: nn.init.constant is now deprecated in favor of nn.init.constant_.
nn.init.constant(m.bias, 0)
Select the networks training parameters¶
lr = 0.0001 # Learning rate
criterion = nn.MSELoss() # Loss function
optim = torch.optim.Adam(network.parameters(), lr=lr) # OptimiserStep four - Network Training¶
Now we have successfully built our network and prepared our data - by patching it to get adequate training samples and creating the input data by selecting and corrupting the active pixels. We are now ready to train the network.
Remember, the network training is slightly different to standard image processing tasks in that we will only be computing the loss on the active pixels.
TO DO: DEFINE TRAINING PARAMETERS¶
# Choose the number of epochs
n_epochs = 100 # most recommend 150-200 for random noise suppression
# Choose number of training and validation samples
n_training = 2048
n_test = 512
# Choose the batch size for the networks training
batch_size = 128# Initialise arrays to keep track of train and validation metrics
train_loss_history = np.zeros(n_epochs)
train_accuracy_history = np.zeros(n_epochs)
test_loss_history = np.zeros(n_epochs)
test_accuracy_history = np.zeros(n_epochs)
# Create torch generator with fixed seed for reproducibility, to be used with the data loaders
g = torch.Generator()
g.manual_seed(0)<torch._C.Generator at 0x2afaace1a5f0>TO DO: INCORPORATE LOSS FUNCTION INTO TRAINING PROCEDURE¶
def n2v_train(model,
criterion,
optimizer,
data_loader,
device):
""" Blind-spot network training function
Parameters
----------
model : torch model
Neural network
criterion : torch criterion
Loss function
optimizer : torch optimizer
Network optimiser
data_loader : torch dataloader
Premade data loader with training data batches
device : torch device
Device where training will occur (e.g., CPU or GPU)
Returns
-------
loss : float
Training loss across full dataset (i.e., all batches)
accuracy : float
Training RMSE accuracy across full dataset (i.e., all batches)
"""
model.train()
accuracy = 0 # initialise accuracy at zero for start of epoch
loss = 0 # initialise loss at zero for start of epoch
for dl in tqdm(data_loader):
# Load batch of data from data loader
X, y, mask = dl[0].to(device), dl[1].to(device), dl[2].to(device)
optimizer.zero_grad()
# Predict the denoised image based on current network weights
yprob = model(X)
# ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
# TO DO: Compute loss function only at masked locations and backpropogate it
# (Hint: only two lines required)
ls = criterion(yprob * (1 - mask), y * (1 - mask))
ls.backward()
# ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
optimizer.step()
with torch.no_grad():
yprob = yprob
ypred = (yprob.detach().cpu().numpy()).astype(float)
# Retain training metrics
loss += ls.item()
accuracy += np.sqrt(np.mean((y.cpu().numpy().ravel( ) - ypred.ravel() )**2))
# Divide cumulative training metrics by number of batches for training
loss /= len(data_loader)
accuracy /= len(data_loader)
return loss, accuracyTO DO: INCORPORATE LOSS FUNCTION INTO VALIDATION PROCEDURE¶
def n2v_evaluate(model,
criterion,
optimizer,
data_loader,
device):
""" Blind-spot network evaluation function
Parameters
----------
model : torch model
Neural network
criterion : torch criterion
Loss function
optimizer : torch optimizer
Network optimiser
data_loader : torch dataloader
Premade data loader with training data batches
device : torch device
Device where network computation will occur (e.g., CPU or GPU)
Returns
-------
loss : float
Validation loss across full dataset (i.e., all batches)
accuracy : float
Validation RMSE accuracy across full dataset (i.e., all batches)
"""
model.train()
accuracy = 0 # initialise accuracy at zero for start of epoch
loss = 0 # initialise loss at zero for start of epoch
for dl in tqdm(data_loader):
# Load batch of data from data loader
X, y, mask = dl[0].to(device), dl[1].to(device), dl[2].to(device)
optimizer.zero_grad()
yprob = model(X)
with torch.no_grad():
# ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
# TO DO: Compute loss function only at masked locations
# (Hint: only one line required)
ls = criterion(yprob * (1 - mask), y * (1 - mask))
# ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
ypred = (yprob.detach().cpu().numpy()).astype(float)
# Retain training metrics
loss += ls.item()
accuracy += np.sqrt(np.mean((y.cpu().numpy().ravel( ) - ypred.ravel() )**2))
# Divide cumulative training metrics by number of batches for training
loss /= len(data_loader)
accuracy /= len(data_loader)
return loss, accuracyTO DO: COMPLETE TRAINING LOOP BY INCORPORATING ABOVE FUNCTIONS¶
# TRAINING
for ep in range(n_epochs):
# RANDOMLY CORRUPT THE NOISY PATCHES
corrupted_patches = np.zeros_like(noisy_patches)
masks = np.zeros_like(corrupted_patches)
for pi in range(len(noisy_patches)):
# TO DO: USE ACTIVE PIXEL FUNCTION TO COMPUTE INPUT DATA AND MASKS
# Hint: One line of code
corrupted_patches[pi], masks[pi] = multi_active_pixels(noisy_patches[pi],
num_activepixels=int(num_activepixels),
neighbourhood_radius=5,
)
# ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
# MAKE DATA LOADERS - using pre-made function
train_loader, test_loader = make_data_loader(noisy_patches,
corrupted_patches,
masks,
n_training,
n_test,
batch_size = batch_size,
torch_generator=g
)
# ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
# TRAIN
# TO DO: Incorporate previously wrote n2v_train function
train_loss, train_accuracy = n2v_train(network,
criterion,
optim,
train_loader,
device,)
# Keeping track of training metrics
train_loss_history[ep], train_accuracy_history[ep] = train_loss, train_accuracy
# ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
# EVALUATE (AKA VALIDATION)
# TO DO: Incorporate previously wrote n2v_evaluate function
test_loss, test_accuracy = n2v_evaluate(network,
criterion,
optim,
test_loader,
device,)
# Keeping track of validation metrics
test_loss_history[ep], test_accuracy_history[ep] = test_loss, test_accuracy
# ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
# PRINTING TRAINING PROGRESS
print(f'''Epoch {ep},
Training Loss {train_loss:.4f}, Training Accuracy {train_accuracy:.4f},
Test Loss {test_loss:.4f}, Test Accuracy {test_accuracy:.4f} ''')
100%|██████████| 16/16 [00:01<00:00, 8.22it/s]
100%|██████████| 4/4 [00:00<00:00, 22.46it/s]
Epoch 0,
Training Loss 0.0003, Training Accuracy 0.1286,
Test Loss 0.0003, Test Accuracy 0.1127
100%|██████████| 16/16 [00:01<00:00, 9.47it/s]
100%|██████████| 4/4 [00:00<00:00, 23.70it/s]
Epoch 1,
Training Loss 0.0002, Training Accuracy 0.1094,
Test Loss 0.0002, Test Accuracy 0.1077
100%|██████████| 16/16 [00:01<00:00, 9.41it/s]
100%|██████████| 4/4 [00:00<00:00, 22.97it/s]
Epoch 2,
Training Loss 0.0002, Training Accuracy 0.1073,
Test Loss 0.0002, Test Accuracy 0.1074
100%|██████████| 16/16 [00:01<00:00, 9.41it/s]
100%|██████████| 4/4 [00:00<00:00, 22.97it/s]
Epoch 3,
Training Loss 0.0002, Training Accuracy 0.1077,
Test Loss 0.0002, Test Accuracy 0.1081
100%|██████████| 16/16 [00:01<00:00, 9.47it/s]
100%|██████████| 4/4 [00:00<00:00, 23.74it/s]
Epoch 4,
Training Loss 0.0002, Training Accuracy 0.1074,
Test Loss 0.0002, Test Accuracy 0.1068
100%|██████████| 16/16 [00:01<00:00, 9.49it/s]
100%|██████████| 4/4 [00:00<00:00, 22.66it/s]
Epoch 5,
Training Loss 0.0002, Training Accuracy 0.1069,
Test Loss 0.0002, Test Accuracy 0.1074
100%|██████████| 16/16 [00:01<00:00, 9.49it/s]
100%|██████████| 4/4 [00:00<00:00, 23.24it/s]
Epoch 6,
Training Loss 0.0002, Training Accuracy 0.1073,
Test Loss 0.0002, Test Accuracy 0.1072
100%|██████████| 16/16 [00:01<00:00, 9.47it/s]
100%|██████████| 4/4 [00:00<00:00, 23.73it/s]
Epoch 7,
Training Loss 0.0002, Training Accuracy 0.1069,
Test Loss 0.0002, Test Accuracy 0.1067
100%|██████████| 16/16 [00:01<00:00, 9.43it/s]
100%|██████████| 4/4 [00:00<00:00, 22.63it/s]
Epoch 8,
Training Loss 0.0002, Training Accuracy 0.1065,
Test Loss 0.0002, Test Accuracy 0.1071
100%|██████████| 16/16 [00:01<00:00, 9.46it/s]
100%|██████████| 4/4 [00:00<00:00, 22.62it/s]
Epoch 9,
Training Loss 0.0002, Training Accuracy 0.1070,
Test Loss 0.0002, Test Accuracy 0.1067
100%|██████████| 16/16 [00:01<00:00, 9.48it/s]
100%|██████████| 4/4 [00:00<00:00, 23.77it/s]
Epoch 10,
Training Loss 0.0002, Training Accuracy 0.1068,
Test Loss 0.0002, Test Accuracy 0.1061
100%|██████████| 16/16 [00:01<00:00, 9.44it/s]
100%|██████████| 4/4 [00:00<00:00, 23.13it/s]
Epoch 11,
Training Loss 0.0002, Training Accuracy 0.1058,
Test Loss 0.0002, Test Accuracy 0.1062
100%|██████████| 16/16 [00:01<00:00, 9.45it/s]
100%|██████████| 4/4 [00:00<00:00, 22.65it/s]
Epoch 12,
Training Loss 0.0002, Training Accuracy 0.1059,
Test Loss 0.0002, Test Accuracy 0.1057
100%|██████████| 16/16 [00:01<00:00, 9.48it/s]
100%|██████████| 4/4 [00:00<00:00, 23.29it/s]
Epoch 13,
Training Loss 0.0002, Training Accuracy 0.1061,
Test Loss 0.0002, Test Accuracy 0.1061
100%|██████████| 16/16 [00:01<00:00, 9.47it/s]
100%|██████████| 4/4 [00:00<00:00, 22.78it/s]
Epoch 14,
Training Loss 0.0002, Training Accuracy 0.1059,
Test Loss 0.0002, Test Accuracy 0.1053
100%|██████████| 16/16 [00:01<00:00, 9.49it/s]
100%|██████████| 4/4 [00:00<00:00, 23.31it/s]
Epoch 15,
Training Loss 0.0002, Training Accuracy 0.1052,
Test Loss 0.0002, Test Accuracy 0.1054
100%|██████████| 16/16 [00:01<00:00, 9.45it/s]
100%|██████████| 4/4 [00:00<00:00, 23.04it/s]
Epoch 16,
Training Loss 0.0002, Training Accuracy 0.1056,
Test Loss 0.0002, Test Accuracy 0.1057
100%|██████████| 16/16 [00:01<00:00, 9.47it/s]
100%|██████████| 4/4 [00:00<00:00, 22.41it/s]
Epoch 17,
Training Loss 0.0002, Training Accuracy 0.1054,
Test Loss 0.0002, Test Accuracy 0.1052
100%|██████████| 16/16 [00:01<00:00, 9.48it/s]
100%|██████████| 4/4 [00:00<00:00, 23.01it/s]
Epoch 18,
Training Loss 0.0002, Training Accuracy 0.1048,
Test Loss 0.0002, Test Accuracy 0.1046
100%|██████████| 16/16 [00:01<00:00, 9.38it/s]
100%|██████████| 4/4 [00:00<00:00, 23.11it/s]
Epoch 19,
Training Loss 0.0002, Training Accuracy 0.1045,
Test Loss 0.0002, Test Accuracy 0.1048
100%|██████████| 16/16 [00:01<00:00, 9.50it/s]
100%|██████████| 4/4 [00:00<00:00, 23.36it/s]
Epoch 20,
Training Loss 0.0002, Training Accuracy 0.1050,
Test Loss 0.0002, Test Accuracy 0.1046
100%|██████████| 16/16 [00:01<00:00, 9.50it/s]
100%|██████████| 4/4 [00:00<00:00, 23.24it/s]
Epoch 21,
Training Loss 0.0002, Training Accuracy 0.1047,
Test Loss 0.0002, Test Accuracy 0.1045
100%|██████████| 16/16 [00:01<00:00, 9.47it/s]
100%|██████████| 4/4 [00:00<00:00, 22.35it/s]
Epoch 22,
Training Loss 0.0002, Training Accuracy 0.1045,
Test Loss 0.0002, Test Accuracy 0.1045
100%|██████████| 16/16 [00:01<00:00, 9.48it/s]
100%|██████████| 4/4 [00:00<00:00, 22.69it/s]
Epoch 23,
Training Loss 0.0002, Training Accuracy 0.1046,
Test Loss 0.0002, Test Accuracy 0.1048
100%|██████████| 16/16 [00:01<00:00, 9.48it/s]
100%|██████████| 4/4 [00:00<00:00, 22.63it/s]
Epoch 24,
Training Loss 0.0002, Training Accuracy 0.1048,
Test Loss 0.0002, Test Accuracy 0.1047
100%|██████████| 16/16 [00:01<00:00, 9.45it/s]
100%|██████████| 4/4 [00:00<00:00, 22.81it/s]
Epoch 25,
Training Loss 0.0002, Training Accuracy 0.1051,
Test Loss 0.0002, Test Accuracy 0.1050
100%|██████████| 16/16 [00:01<00:00, 9.48it/s]
100%|██████████| 4/4 [00:00<00:00, 22.44it/s]
Epoch 26,
Training Loss 0.0002, Training Accuracy 0.1047,
Test Loss 0.0002, Test Accuracy 0.1044
100%|██████████| 16/16 [00:01<00:00, 9.49it/s]
100%|██████████| 4/4 [00:00<00:00, 22.65it/s]
Epoch 27,
Training Loss 0.0002, Training Accuracy 0.1041,
Test Loss 0.0002, Test Accuracy 0.1039
100%|██████████| 16/16 [00:01<00:00, 9.47it/s]
100%|██████████| 4/4 [00:00<00:00, 23.23it/s]
Epoch 28,
Training Loss 0.0002, Training Accuracy 0.1038,
Test Loss 0.0002, Test Accuracy 0.1036
100%|██████████| 16/16 [00:01<00:00, 9.48it/s]
100%|██████████| 4/4 [00:00<00:00, 22.64it/s]
Epoch 29,
Training Loss 0.0002, Training Accuracy 0.1039,
Test Loss 0.0002, Test Accuracy 0.1041
100%|██████████| 16/16 [00:01<00:00, 9.47it/s]
100%|██████████| 4/4 [00:00<00:00, 23.24it/s]
Epoch 30,
Training Loss 0.0002, Training Accuracy 0.1038,
Test Loss 0.0002, Test Accuracy 0.1040
100%|██████████| 16/16 [00:01<00:00, 9.44it/s]
100%|██████████| 4/4 [00:00<00:00, 22.63it/s]
Epoch 31,
Training Loss 0.0002, Training Accuracy 0.1041,
Test Loss 0.0002, Test Accuracy 0.1044
100%|██████████| 16/16 [00:01<00:00, 9.45it/s]
100%|██████████| 4/4 [00:00<00:00, 23.27it/s]
Epoch 32,
Training Loss 0.0002, Training Accuracy 0.1041,
Test Loss 0.0002, Test Accuracy 0.1041
100%|██████████| 16/16 [00:01<00:00, 9.46it/s]
100%|██████████| 4/4 [00:00<00:00, 23.06it/s]
Epoch 33,
Training Loss 0.0002, Training Accuracy 0.1041,
Test Loss 0.0002, Test Accuracy 0.1038
100%|██████████| 16/16 [00:01<00:00, 9.48it/s]
100%|██████████| 4/4 [00:00<00:00, 23.29it/s]
Epoch 34,
Training Loss 0.0002, Training Accuracy 0.1037,
Test Loss 0.0002, Test Accuracy 0.1035
100%|██████████| 16/16 [00:01<00:00, 9.52it/s]
100%|██████████| 4/4 [00:00<00:00, 22.72it/s]
Epoch 35,
Training Loss 0.0002, Training Accuracy 0.1036,
Test Loss 0.0002, Test Accuracy 0.1037
100%|██████████| 16/16 [00:01<00:00, 9.47it/s]
100%|██████████| 4/4 [00:00<00:00, 23.29it/s]
Epoch 36,
Training Loss 0.0002, Training Accuracy 0.1042,
Test Loss 0.0002, Test Accuracy 0.1044
100%|██████████| 16/16 [00:01<00:00, 9.46it/s]
100%|██████████| 4/4 [00:00<00:00, 23.26it/s]
Epoch 37,
Training Loss 0.0002, Training Accuracy 0.1041,
Test Loss 0.0002, Test Accuracy 0.1038
100%|██████████| 16/16 [00:01<00:00, 9.49it/s]
100%|██████████| 4/4 [00:00<00:00, 23.26it/s]
Epoch 38,
Training Loss 0.0002, Training Accuracy 0.1034,
Test Loss 0.0002, Test Accuracy 0.1035
100%|██████████| 16/16 [00:01<00:00, 9.47it/s]
100%|██████████| 4/4 [00:00<00:00, 22.62it/s]
Epoch 39,
Training Loss 0.0002, Training Accuracy 0.1033,
Test Loss 0.0002, Test Accuracy 0.1034
100%|██████████| 16/16 [00:01<00:00, 9.41it/s]
100%|██████████| 4/4 [00:00<00:00, 23.03it/s]
Epoch 40,
Training Loss 0.0002, Training Accuracy 0.1031,
Test Loss 0.0002, Test Accuracy 0.1032
100%|██████████| 16/16 [00:01<00:00, 9.48it/s]
100%|██████████| 4/4 [00:00<00:00, 23.25it/s]
Epoch 41,
Training Loss 0.0002, Training Accuracy 0.1033,
Test Loss 0.0002, Test Accuracy 0.1037
100%|██████████| 16/16 [00:01<00:00, 9.47it/s]
100%|██████████| 4/4 [00:00<00:00, 23.15it/s]
Epoch 42,
Training Loss 0.0002, Training Accuracy 0.1035,
Test Loss 0.0002, Test Accuracy 0.1037
100%|██████████| 16/16 [00:01<00:00, 9.46it/s]
100%|██████████| 4/4 [00:00<00:00, 23.96it/s]
Epoch 43,
Training Loss 0.0002, Training Accuracy 0.1039,
Test Loss 0.0002, Test Accuracy 0.1039
100%|██████████| 16/16 [00:01<00:00, 9.50it/s]
100%|██████████| 4/4 [00:00<00:00, 23.35it/s]
Epoch 44,
Training Loss 0.0002, Training Accuracy 0.1036,
Test Loss 0.0002, Test Accuracy 0.1032
100%|██████████| 16/16 [00:01<00:00, 9.49it/s]
100%|██████████| 4/4 [00:00<00:00, 23.23it/s]
Epoch 45,
Training Loss 0.0002, Training Accuracy 0.1035,
Test Loss 0.0002, Test Accuracy 0.1034
100%|██████████| 16/16 [00:01<00:00, 9.46it/s]
100%|██████████| 4/4 [00:00<00:00, 23.24it/s]
Epoch 46,
Training Loss 0.0002, Training Accuracy 0.1034,
Test Loss 0.0002, Test Accuracy 0.1035
100%|██████████| 16/16 [00:01<00:00, 9.46it/s]
100%|██████████| 4/4 [00:00<00:00, 23.09it/s]
Epoch 47,
Training Loss 0.0002, Training Accuracy 0.1038,
Test Loss 0.0002, Test Accuracy 0.1042
100%|██████████| 16/16 [00:01<00:00, 9.49it/s]
100%|██████████| 4/4 [00:00<00:00, 22.67it/s]
Epoch 48,
Training Loss 0.0002, Training Accuracy 0.1038,
Test Loss 0.0002, Test Accuracy 0.1035
100%|██████████| 16/16 [00:01<00:00, 9.47it/s]
100%|██████████| 4/4 [00:00<00:00, 23.25it/s]
Epoch 49,
Training Loss 0.0002, Training Accuracy 0.1034,
Test Loss 0.0002, Test Accuracy 0.1037
100%|██████████| 16/16 [00:01<00:00, 9.40it/s]
100%|██████████| 4/4 [00:00<00:00, 22.95it/s]
Epoch 50,
Training Loss 0.0002, Training Accuracy 0.1034,
Test Loss 0.0002, Test Accuracy 0.1031
100%|██████████| 16/16 [00:01<00:00, 9.44it/s]
100%|██████████| 4/4 [00:00<00:00, 23.24it/s]
Epoch 51,
Training Loss 0.0002, Training Accuracy 0.1031,
Test Loss 0.0002, Test Accuracy 0.1030
100%|██████████| 16/16 [00:01<00:00, 9.48it/s]
100%|██████████| 4/4 [00:00<00:00, 22.30it/s]
Epoch 52,
Training Loss 0.0002, Training Accuracy 0.1033,
Test Loss 0.0002, Test Accuracy 0.1035
100%|██████████| 16/16 [00:01<00:00, 9.47it/s]
100%|██████████| 4/4 [00:00<00:00, 23.38it/s]
Epoch 53,
Training Loss 0.0002, Training Accuracy 0.1033,
Test Loss 0.0002, Test Accuracy 0.1029
100%|██████████| 16/16 [00:01<00:00, 9.45it/s]
100%|██████████| 4/4 [00:00<00:00, 23.07it/s]
Epoch 54,
Training Loss 0.0002, Training Accuracy 0.1031,
Test Loss 0.0002, Test Accuracy 0.1035
100%|██████████| 16/16 [00:01<00:00, 9.48it/s]
100%|██████████| 4/4 [00:00<00:00, 23.25it/s]
Epoch 55,
Training Loss 0.0002, Training Accuracy 0.1033,
Test Loss 0.0002, Test Accuracy 0.1032
100%|██████████| 16/16 [00:01<00:00, 9.44it/s]
100%|██████████| 4/4 [00:00<00:00, 23.26it/s]
Epoch 56,
Training Loss 0.0002, Training Accuracy 0.1034,
Test Loss 0.0002, Test Accuracy 0.1036
100%|██████████| 16/16 [00:01<00:00, 9.47it/s]
100%|██████████| 4/4 [00:00<00:00, 22.97it/s]
Epoch 57,
Training Loss 0.0002, Training Accuracy 0.1035,
Test Loss 0.0002, Test Accuracy 0.1031
100%|██████████| 16/16 [00:01<00:00, 9.48it/s]
100%|██████████| 4/4 [00:00<00:00, 22.56it/s]
Epoch 58,
Training Loss 0.0002, Training Accuracy 0.1030,
Test Loss 0.0002, Test Accuracy 0.1030
100%|██████████| 16/16 [00:01<00:00, 9.54it/s]
100%|██████████| 4/4 [00:00<00:00, 22.43it/s]
Epoch 59,
Training Loss 0.0002, Training Accuracy 0.1032,
Test Loss 0.0002, Test Accuracy 0.1033
100%|██████████| 16/16 [00:01<00:00, 9.54it/s]
100%|██████████| 4/4 [00:00<00:00, 23.25it/s]
Epoch 60,
Training Loss 0.0002, Training Accuracy 0.1032,
Test Loss 0.0002, Test Accuracy 0.1031
100%|██████████| 16/16 [00:01<00:00, 9.47it/s]
100%|██████████| 4/4 [00:00<00:00, 23.10it/s]
Epoch 61,
Training Loss 0.0002, Training Accuracy 0.1033,
Test Loss 0.0002, Test Accuracy 0.1032
100%|██████████| 16/16 [00:01<00:00, 9.47it/s]
100%|██████████| 4/4 [00:00<00:00, 22.94it/s]
Epoch 62,
Training Loss 0.0002, Training Accuracy 0.1030,
Test Loss 0.0002, Test Accuracy 0.1029
100%|██████████| 16/16 [00:01<00:00, 9.48it/s]
100%|██████████| 4/4 [00:00<00:00, 22.62it/s]
Epoch 63,
Training Loss 0.0002, Training Accuracy 0.1032,
Test Loss 0.0002, Test Accuracy 0.1034
100%|██████████| 16/16 [00:01<00:00, 9.42it/s]
100%|██████████| 4/4 [00:00<00:00, 23.24it/s]
Epoch 64,
Training Loss 0.0002, Training Accuracy 0.1032,
Test Loss 0.0002, Test Accuracy 0.1029
100%|██████████| 16/16 [00:01<00:00, 9.49it/s]
100%|██████████| 4/4 [00:00<00:00, 23.21it/s]
Epoch 65,
Training Loss 0.0002, Training Accuracy 0.1029,
Test Loss 0.0002, Test Accuracy 0.1034
100%|██████████| 16/16 [00:01<00:00, 9.43it/s]
100%|██████████| 4/4 [00:00<00:00, 23.26it/s]
Epoch 66,
Training Loss 0.0002, Training Accuracy 0.1034,
Test Loss 0.0002, Test Accuracy 0.1033
100%|██████████| 16/16 [00:01<00:00, 9.46it/s]
100%|██████████| 4/4 [00:00<00:00, 23.08it/s]
Epoch 67,
Training Loss 0.0002, Training Accuracy 0.1026,
Test Loss 0.0002, Test Accuracy 0.1026
100%|██████████| 16/16 [00:01<00:00, 9.48it/s]
100%|██████████| 4/4 [00:00<00:00, 23.30it/s]
Epoch 68,
Training Loss 0.0002, Training Accuracy 0.1028,
Test Loss 0.0002, Test Accuracy 0.1034
100%|██████████| 16/16 [00:01<00:00, 9.49it/s]
100%|██████████| 4/4 [00:00<00:00, 23.35it/s]
Epoch 69,
Training Loss 0.0002, Training Accuracy 0.1031,
Test Loss 0.0002, Test Accuracy 0.1032
100%|██████████| 16/16 [00:01<00:00, 9.43it/s]
100%|██████████| 4/4 [00:00<00:00, 23.35it/s]
Epoch 70,
Training Loss 0.0002, Training Accuracy 0.1032,
Test Loss 0.0002, Test Accuracy 0.1031
100%|██████████| 16/16 [00:01<00:00, 9.43it/s]
100%|██████████| 4/4 [00:00<00:00, 23.68it/s]
Epoch 71,
Training Loss 0.0002, Training Accuracy 0.1031,
Test Loss 0.0002, Test Accuracy 0.1031
100%|██████████| 16/16 [00:01<00:00, 9.46it/s]
100%|██████████| 4/4 [00:00<00:00, 22.39it/s]
Epoch 72,
Training Loss 0.0002, Training Accuracy 0.1032,
Test Loss 0.0002, Test Accuracy 0.1028
100%|██████████| 16/16 [00:01<00:00, 9.48it/s]
100%|██████████| 4/4 [00:00<00:00, 23.39it/s]
Epoch 73,
Training Loss 0.0002, Training Accuracy 0.1026,
Test Loss 0.0002, Test Accuracy 0.1028
100%|██████████| 16/16 [00:01<00:00, 9.49it/s]
100%|██████████| 4/4 [00:00<00:00, 23.24it/s]
Epoch 74,
Training Loss 0.0002, Training Accuracy 0.1029,
Test Loss 0.0002, Test Accuracy 0.1032
100%|██████████| 16/16 [00:01<00:00, 9.46it/s]
100%|██████████| 4/4 [00:00<00:00, 23.04it/s]
Epoch 75,
Training Loss 0.0002, Training Accuracy 0.1030,
Test Loss 0.0002, Test Accuracy 0.1034
100%|██████████| 16/16 [00:01<00:00, 9.46it/s]
100%|██████████| 4/4 [00:00<00:00, 22.65it/s]
Epoch 76,
Training Loss 0.0002, Training Accuracy 0.1032,
Test Loss 0.0002, Test Accuracy 0.1032
100%|██████████| 16/16 [00:01<00:00, 9.48it/s]
100%|██████████| 4/4 [00:00<00:00, 22.60it/s]
Epoch 77,
Training Loss 0.0002, Training Accuracy 0.1032,
Test Loss 0.0002, Test Accuracy 0.1034
100%|██████████| 16/16 [00:01<00:00, 9.47it/s]
100%|██████████| 4/4 [00:00<00:00, 22.71it/s]
Epoch 78,
Training Loss 0.0002, Training Accuracy 0.1032,
Test Loss 0.0002, Test Accuracy 0.1028
100%|██████████| 16/16 [00:01<00:00, 9.51it/s]
100%|██████████| 4/4 [00:00<00:00, 23.21it/s]
Epoch 79,
Training Loss 0.0002, Training Accuracy 0.1026,
Test Loss 0.0002, Test Accuracy 0.1027
100%|██████████| 16/16 [00:01<00:00, 9.49it/s]
100%|██████████| 4/4 [00:00<00:00, 23.27it/s]
Epoch 80,
Training Loss 0.0002, Training Accuracy 0.1024,
Test Loss 0.0002, Test Accuracy 0.1026
100%|██████████| 16/16 [00:01<00:00, 9.44it/s]
100%|██████████| 4/4 [00:00<00:00, 23.24it/s]
Epoch 81,
Training Loss 0.0002, Training Accuracy 0.1026,
Test Loss 0.0002, Test Accuracy 0.1031
100%|██████████| 16/16 [00:01<00:00, 9.43it/s]
100%|██████████| 4/4 [00:00<00:00, 22.66it/s]
Epoch 82,
Training Loss 0.0002, Training Accuracy 0.1031,
Test Loss 0.0002, Test Accuracy 0.1034
100%|██████████| 16/16 [00:01<00:00, 9.47it/s]
100%|██████████| 4/4 [00:00<00:00, 22.41it/s]
Epoch 83,
Training Loss 0.0002, Training Accuracy 0.1032,
Test Loss 0.0002, Test Accuracy 0.1032
100%|██████████| 16/16 [00:01<00:00, 9.51it/s]
100%|██████████| 4/4 [00:00<00:00, 22.65it/s]
Epoch 84,
Training Loss 0.0002, Training Accuracy 0.1027,
Test Loss 0.0002, Test Accuracy 0.1025
100%|██████████| 16/16 [00:01<00:00, 9.51it/s]
100%|██████████| 4/4 [00:00<00:00, 23.82it/s]
Epoch 85,
Training Loss 0.0002, Training Accuracy 0.1026,
Test Loss 0.0002, Test Accuracy 0.1030
100%|██████████| 16/16 [00:01<00:00, 9.44it/s]
100%|██████████| 4/4 [00:00<00:00, 23.24it/s]
Epoch 86,
Training Loss 0.0002, Training Accuracy 0.1027,
Test Loss 0.0002, Test Accuracy 0.1025
100%|██████████| 16/16 [00:01<00:00, 9.48it/s]
100%|██████████| 4/4 [00:00<00:00, 22.68it/s]
Epoch 87,
Training Loss 0.0002, Training Accuracy 0.1024,
Test Loss 0.0002, Test Accuracy 0.1027
100%|██████████| 16/16 [00:01<00:00, 9.49it/s]
100%|██████████| 4/4 [00:00<00:00, 23.81it/s]
Epoch 88,
Training Loss 0.0002, Training Accuracy 0.1030,
Test Loss 0.0002, Test Accuracy 0.1030
100%|██████████| 16/16 [00:01<00:00, 9.41it/s]
100%|██████████| 4/4 [00:00<00:00, 23.25it/s]
Epoch 89,
Training Loss 0.0002, Training Accuracy 0.1027,
Test Loss 0.0002, Test Accuracy 0.1024
100%|██████████| 16/16 [00:01<00:00, 9.48it/s]
100%|██████████| 4/4 [00:00<00:00, 22.75it/s]
Epoch 90,
Training Loss 0.0002, Training Accuracy 0.1027,
Test Loss 0.0002, Test Accuracy 0.1028
100%|██████████| 16/16 [00:01<00:00, 9.43it/s]
100%|██████████| 4/4 [00:00<00:00, 23.80it/s]
Epoch 91,
Training Loss 0.0002, Training Accuracy 0.1026,
Test Loss 0.0002, Test Accuracy 0.1023
100%|██████████| 16/16 [00:01<00:00, 9.43it/s]
100%|██████████| 4/4 [00:00<00:00, 23.18it/s]
Epoch 92,
Training Loss 0.0002, Training Accuracy 0.1022,
Test Loss 0.0002, Test Accuracy 0.1025
100%|██████████| 16/16 [00:01<00:00, 9.47it/s]
100%|██████████| 4/4 [00:00<00:00, 23.40it/s]
Epoch 93,
Training Loss 0.0002, Training Accuracy 0.1027,
Test Loss 0.0002, Test Accuracy 0.1030
100%|██████████| 16/16 [00:01<00:00, 9.49it/s]
100%|██████████| 4/4 [00:00<00:00, 23.27it/s]
Epoch 94,
Training Loss 0.0002, Training Accuracy 0.1027,
Test Loss 0.0002, Test Accuracy 0.1025
100%|██████████| 16/16 [00:01<00:00, 9.43it/s]
100%|██████████| 4/4 [00:00<00:00, 23.25it/s]
Epoch 95,
Training Loss 0.0002, Training Accuracy 0.1023,
Test Loss 0.0002, Test Accuracy 0.1026
100%|██████████| 16/16 [00:01<00:00, 9.48it/s]
100%|██████████| 4/4 [00:00<00:00, 23.08it/s]
Epoch 96,
Training Loss 0.0002, Training Accuracy 0.1030,
Test Loss 0.0002, Test Accuracy 0.1030
100%|██████████| 16/16 [00:01<00:00, 9.48it/s]
100%|██████████| 4/4 [00:00<00:00, 22.38it/s]
Epoch 97,
Training Loss 0.0002, Training Accuracy 0.1027,
Test Loss 0.0002, Test Accuracy 0.1024
100%|██████████| 16/16 [00:01<00:00, 9.48it/s]
100%|██████████| 4/4 [00:00<00:00, 22.67it/s]
Epoch 98,
Training Loss 0.0002, Training Accuracy 0.1026,
Test Loss 0.0002, Test Accuracy 0.1027
100%|██████████| 16/16 [00:01<00:00, 9.50it/s]
100%|██████████| 4/4 [00:00<00:00, 22.59it/s]Epoch 99,
Training Loss 0.0002, Training Accuracy 0.1027,
Test Loss 0.0002, Test Accuracy 0.1026
# Plotting trainnig metrics using pre-made function
fig,axs = plot_training_metrics(train_accuracy_history,
test_accuracy_history,
train_loss_history,
test_loss_history
)
Step five - Apply trained model¶
The model is now trained and ready for its denoising capabilities to be tested.
For the standard network application, the noisy image does not require any data patching nor does it require the active pixel pre-processing required in training. In other words, the noisy image can be fed directly into the network for denoising.
TO DO: DENOISE NEW NOISY DATASET¶
# Make a new noisy realisation so it's different from the training set but with roughly same level of noise
testdata, _ = add_whitegaussian_noise(d, sc=0.1)
# Convert dataset in tensor for prediction purposes
torch_testdata = torch.from_numpy(np.expand_dims(np.expand_dims(testdata,axis=0),axis=0)).float()
# Run test dataset through network
network.eval()
test_prediction = network(torch_testdata.to(device))
# Return to numpy for plotting purposes
test_pred = test_prediction.detach().cpu().numpy().squeeze()
# Use pre-made plotting function to visualise denoising performance
fig,axs = plot_synth_results(d, testdata, test_pred)