%load_ext autoreload
%autoreload 2

os.environ["CUDA_VISIBLE_DEVICES"] = '0'

path_tif = Path('/groups/turaga/home/speisera/share_TUM/FishSIM/sim_1/mRNAlevel_200/cell3D/strong/w1_HelaKyoto_Gapdh_2597_p01_cy3__Cell_CP_14__cell3D__1.tif')
bead_vol = load_tiff_image(path_tif)
estimate_backg = EstimateBackground(20, 1)

bead_vol = bead_vol - estimate_backg(bead_vol)

win_sl = np.s_[0,:,50:95,75:160]

plt.figure(figsize=(20,6))
plt.subplot(131)
plt.imshow(bead_vol[0].mean(0))
plt.subplot(132)
plt.imshow(bead_vol[win_sl].max(0).values)
plt.subplot(133)
plt.imshow(bead_vol[win_sl].mean(1))

<matplotlib.image.AxesImage at 0x7fceebdf2ad0>

get_peaks_3d[source]

get_peaks_3d(volume, threshold=500, min_distance=20)

Peak finding functions. Provides position estimate for bead stacks that are used as initialization for PSF fitting.

Parameters
----------
volume: 3D array
    Single bead recording
threshold: float
    Initial threshold to identify pixel that are considered as possible peaks.
min_distance: float
    Minimal distance between two peaks in pixels

Returns
-------
peaks: array
    Array of z,y,x peak positions

plot_detection[source]

plot_detection(volume, coordinates)

coords_xyz = get_peaks_3d(bead_vol[win_sl], threshold=2000, min_distance=5)
plot_detection(bead_vol[win_sl], coords_xyz)

abs_coords_xyz = np.array([c + np.array([int(win_sl[3].start or 0), int(win_sl[2].start or 0), int(win_sl[1].start or 0)]) for c in coords_xyz])

extract_roi[source]

extract_roi(beads:Tensor, coordinates:Tensor, size_xy:int=10, size_z:int=10)

rois = extract_roi(bead_vol, abs_coords_xyz, size_xy=10, size_z=10)

for i in range(len(rois)): plot_3d_projections(rois[i], projection='max', size=3);

print(rois.shape)
rois_clean = rois[[0,2,4,5]]
rois_normed = rois_clean/rois_clean.sum(-1).sum(-1).sum(-1)[:,None,None,None] * 100

extend = [0,5,5,5]
rois_extended = torch.zeros([s + 2*e for s,e in zip(rois_normed.shape,extend)])
rois_extended[:,extend[1]:-extend[1],extend[2]:-extend[2],extend[3]:-extend[3]] = rois_normed

torch.Size([6, 21, 21, 21])

device='cuda'
PSF = LinearInterpolatedPSF(rois_extended.shape[-3:],1, device=device)

fit_psf[source]

fit_psf(PSF, roi_list, fit_lr=0.001, fit_iteration=10000, device='cuda')

loss_res = fit_psf(PSF, rois_extended)

plot_3d_projections(PSF.psf_volume[0], projection='mean', size=3);

plot_3d_projections(torch.clamp_min(PSF.psf_volume[0],0), projection='mean', size=3);

plt.plot(loss_res)
plt.yscale('log')
plt.plot([0,10000],[np.min(loss_res),np.min(loss_res)])
plt.title(np.min(loss_res))

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
<ipython-input-37-82fd6e63650e> in <module>
----> 1 plt.plot(loss_res)
      2 plt.yscale('log')
      3 plt.plot([0,10000],[np.min(loss_res),np.min(loss_res)])
      4 plt.title(np.min(loss_res))

NameError: name 'loss_res' is not defined

PSF.psf_volume.data = torch.clamp_min_(PSF.psf_volume.data, 0) 

torch.save(PSF.state_dict(), '../data/simfish_psf.pkl')

from scipy import ndimage
delta = torch.zeros_like(PSF.psf_volume.detach())
delta[0,15,15,15] = 1
gauss_3d = 100*scipy.ndimage.filters.gaussian_filter(delta[0], 1.2, order=0, output=None, mode='reflect', cval=0.0, truncate=4.0)
plot_3d_projections(gauss_3d)

gt_psf = load_tiff_image('/groups/turaga/home/speisera/share_TUM/FishSIM/PSF.tif')
plot_3d_projections(gt_psf[0,::3,::3,::3])

device='cuda'
PSF = LinearInterpolatedPSF([21,21,21],1, device=device)

PSF.psf_volume.data = torch.tensor(gauss_3d[None])
torch.save(PSF.state_dict(), '../data/gaussian_psf.pkl')

!nbdev_build_lib

Converted 00_models.ipynb.
Converted 01_psf.ipynb.
Converted 02_microscope.ipynb.
Converted 03_noise.ipynb.
Converted 04_pointsource.ipynb.
Converted 05_gmm_loss.ipynb.
Converted 06_plotting.ipynb.
Converted 07_file_io.ipynb.
Converted 08_dataset.ipynb.
Converted 09_output_trafo.ipynb.
Converted 10_evaluation.ipynb.
Converted 11_emitter_io.ipynb.
Converted 12_utils.ipynb.
Converted 13_train.ipynb.
Converted 15_fit_psf.ipynb.
Converted 16_visualization.ipynb.
Converted index.ipynb.