ax[1].set_title('Trainset error = {:.2e}'.format(trainset_error), fontsize=fs)
ax[1].axes.xaxis.set_ticks([]) #set_xticklabels([])
ax[1].axes.yaxis.set_ticks([]) #set_yticklabels([])
ax[1].set_xlabel('$e$ = {:.2e} PSNR = {:.2f}'.format(trainimage_error,
psnrscore),
fontsize=fs)
filename = "Just Generator/minerrortest" + timestamp + ".png"
fig.savefig(filename, bbox_inches='tight')
trainimages = testTheseImages(train_LRimages, model)
print(train_HRimages[0].shape)
trainset_error = sum([
np.mean(np.sqrt((i - j)**2)) for i, j in zip(trainimages, train_HRimages)
]) / len(trainimages)
print(trainset_error)
trainset_error = sum(
[lf.smooth_l1(i, j)
for i, j in zip(trainimages, train_HRimages)]) / len(trainimages)
print(trainset_error)
testimages = testTheseImages(test_LRimages, model)
testset_error = sum(
[np.mean(np.sqrt((i - j)**2))
for i, j in zip(testimages, test_HRimages)]) / len(testimages)
print(testset_error)
testset_error = sum(
[lf.smooth_l1(i, j)
for i, j in zip(testimages, test_HRimages)]) / len(testimages)
print(testset_error)
testindex = 4
ax[2].imshow(test_HRimages[testindex], cmap='gray')
ax[2].set_title("Input from test set", fontsize=fs)
ax[2].axis('off')
testimage = testThisImage(test_LRimages[testindex], model)
testimage_error = np.mean((testimage - test_HRimages[testindex])**2)
psnrscore = psnr(test_HRimages[testindex].astype(np.float),
testimage.astype(np.float))
testimages = testTheseImages(test_LRimages, model)
testset_error = sum(
[np.mean((i - j)**2)
for i, j in zip(testimages, test_HRimages)]) / len(testimages)
print(testset_error)
print(
sum([lf.smooth_l1(i, j)
for i, j in zip(testimages, test_HRimages)]) / len(testimages))
ax[3].imshow(testimage, cmap='gray')
ax[3].set_title('Testset error = {:.2e}'.format(testset_error), fontsize=fs)
ax[3].axes.xaxis.set_ticks([]) #set_xticklabels([])
ax[3].axes.yaxis.set_ticks([]) #set_yticklabels([])
ax[3].set_xlabel('$e$ = {:.2e} PSNR = {:.2f}'.format(testimage_error,
psnrscore),
fontsize=fs)
# average = torch.zeros(HR_dim, HR_dim)
# for HiResIm, LoResIm in zip(HR_loader, LR_loader):
# average += HiResIm.mean(0)
fig.subplots_adjust(wspace=0.01, hspace=0.01)