Missing librairies when running demo5_train notebook

[abdelneuhaus]

Hello,

I have been using uiPSF method developed by Jonas Ries lab and at some point they said to use FD-DeepLoc to do the training for localization analysis using their h5 file output.

In their case, they use specific librairies (such as omegaconf) and an io folder containing function to read h5 files. In your case (I'm refering to demo5_train notebook), you are calling the same io folder from local_utils but it is missing, as well as the omegaconf dependency.

I don't know if I need to "merge" both environment dependencies or code parts from FD-DeepLoc and uiPSF? Or create a new environment with updated librairies?

Thank you for your help,
AN

[abdelneuhaus]

Hello, this issue has been resolved from my side. But I have another one. Whenever I use the demo5_train notebook, I have a bug when this function is called : look_trainingdata(), with this error:

`---------------------------------------------------------------------------
IndexError Traceback (most recent call last)
Cell In[6], line 2
1 # check the training data, camera noise, etc.
----> 2 model.look_trainingdata(area_num=1)
3 # model.look_trainingdata(area_num=181)

File ~\Desktop\Field Dependent PSF Learning\demo_notebooks\demo5_FD_in-situ_astig_NPC../..\fd_deeploc_core\train_loss_rec.py:71, in TrainFuncs.look_trainingdata(self, area_num)
65 map[0, int(self.dat_generator.simulation_pars['margin_empty'] * train_size):
66 int((1 - self.dat_generator.simulation_pars['margin_empty']) * train_size),
67 int(self.dat_generator.simulation_pars['margin_empty'] * train_size):
68 int((1 - self.dat_generator.simulation_pars['margin_empty']) * train_size)] += 1
69 map = map / map.sum() * density
---> 71 imgs_sim, xyzi_gt, s_mask, bg, locs, field_xy = self.dat_generator.sim_func(
72 M=gpu(map), batch_size=1,
73 local_context=self.local_context,
74 photon_filt=False,
75 photon_filt_thre=0,
76 P_locs_cse=True,
77 iter_num=area_num, train_size=train_size,
78 robust_training=self.dat_generator.simulation_pars['robust_training'])
80 bg_photons = (self.dat_generator.simulation_pars['backg'] - self.dat_generator.simulation_pars['baseline'])
81 / self.dat_generator.simulation_pars['em_gain'] * self.dat_generator.simulation_pars['e_per_adu']
82 / self.dat_generator.simulation_pars['qe']
84 mol_photons = (self.dat_generator.simulation_pars['min_ph'] + 1) / 2 * self.dat_generator.psf_pars['ph_scale']

File ~\Desktop\Field Dependent PSF Learning\demo_notebooks\demo5_FD_in-situ_astig_NPC../..\fd_deeploc_core\data_simulator.py:330, in DataSimulator.sim_func(self, M, batch_size, local_context, photon_filt, photon_filt_thre, P_locs_cse, iter_num, train_size, robust_training)
325 S, X_os, Y_os, Z, I, field_xy = self.sampling(batch_size=batch_size, M=M, local_context=local_context,
326 iter_num=iter_num, train_size=train_size)
328 if S.sum():
--> 330 imgs_sim += self.datagen_func(S, X_os, Y_os, Z, I, field_xy, robust_training)
331 xyzi = torch.cat([X_os[:, :, None], Y_os[:, :, None], Z[:, :, None], I[:, :, None]], 2)
333 S = S[:, 1] if local_context else S[:, 0]

File ~\Desktop\Field Dependent PSF Learning\demo_notebooks\demo5_FD_in-situ_astig_NPC../..\fd_deeploc_core\data_simulator.py:149, in DataSimulator.datagen_func(self, S, X_os, Y_os, Z, I, field_xy, robust_training)
146 batch_size, n_inp, h, w = S.shape[0], S.shape[1], S.shape[2], S.shape[3]
147 xyzi = torch.cat([X_os.reshape([-1, 1, h, w]), Y_os.reshape([-1, 1, h, w]), Z.reshape([-1, 1, h, w]),
148 I.reshape([-1, 1, h, w])], 1)
--> 149 recs = self.genfunc(S.reshape([-1, h, w]), xyzi, field_xy, robust_training)
150 torch.clamp_min_(recs, 0)
151 imgs_sim = recs.reshape([batch_size, n_inp, h, w])

File ~\Desktop\Field Dependent PSF Learning\demo_notebooks\demo5_FD_in-situ_astig_NPC../..\fd_deeploc_core\data_simulator.py:142, in DataSimulator.genfunc(self, S, XYZI, field_xy, robust_training)
140 def genfunc(self, S, XYZI, field_xy, robust_training):
141 X_os, Y_os, Z, I = self.transform_offsets(S, XYZI)
--> 142 psf_patches = self.psf_vector(S, X_os, Y_os, Z, I, field_xy, robust_training)
143 return self.psf_pars['ph_scale'] * self.place_psfs(psf_patches, S)

File ~\Desktop\Field Dependent PSF Learning\demo_notebooks\demo5_FD_in-situ_astig_NPC../..\fd_deeploc_core\data_simulator.py:34, in DataSimulator.psf_vector(self, S, X_os, Y_os, Z, I, field_xy, robust_training)
33 def psf_vector(self, S, X_os, Y_os, Z, I, field_xy, robust_training):
---> 34 W = self.sim_psf_vector(cpu(S.nonzero()), cpu(X_os), cpu(Y_os), cpu(Z), self.aber_map,
35 field_xy, robust_training)
37 W /= W.sum(-1).sum(-1)[:, None, None] # photon normalization
38 W *= I

File ~\Desktop\Field Dependent PSF Learning\demo_notebooks\demo5_FD_in-situ_astig_NPC../..\PSF_vector_gpu_PSF_vector_gpu.py:152, in PSF_VECTOR_GPU.sim_psf_vector(self, S, X_os, Y_os, Z, aber_map, field_xy, robust_training)
150 aberrationsImgParas[n] = aber_map_crop[S[n, 1], S[n, 2], :]
151 else:
--> 152 aberrationsRealParas[n] = aber_map_crop[S[n, 1], S[n, 2], :, 0]
153 aberrationsImgParas[n] = aber_map_crop[S[n, 1], S[n, 2], :, 1]
155 # aberrationsParas = aberrationsParas * sP.lambdaX_ # multiply with lambda
156 # sP.aberrationsParas_ = aberrationsParas.ctypes.data_as(POINTER(c_float))

IndexError: index 50 is out of bounds for axis 0 with size 47`

Do you know where it comes from ?

AN

[fffffsssss]

Hi, I am very very sorry for this too-late reply. I haven't opened this repo for quite a long time. From the error message, it seems that the incompatibility of Zernike aberrations between uipsf and fd-deeploc caused this. Please check this, @JianweiChen8