Example scripts results¶
On this page of the documentation we show images of results one can expect to obtain when running the code examples provided with this repo.
In examples are various example scripts that showcase the main features
of this repository.
First, activate the virtual environment:
source mask_designer_env/bin/activate
You can exit the virtual environment by running deactivate.
Lens examples¶
GS¶
python examples/lens/run_gs.py
The simulated results should be similar to the following:
And the corresponding captured image should look something like:
Holoeye¶
python examples/lens/run_holoeye.py
The simulated results should be similar to the following:
And the corresponding captured image should look something like:
SGD using CITL¶
python examples/lens/run_sgd_citl.py
Here is the simulated result of the “warm start” phase which only runs normal SGD.
And here is the captured image of the resulting amplitude of the “warm start” phase.
For improving the results, we capture the 0th order refraction by simply program
a blank phase mask to the SLM and capturing the resulting amplitude. This is
then set as corrective image inside the Camera class, and hence subtracted
from all subsequent images that are taken with the camera.
Hence, when programming the phase mask corresponding to the grid used for calibration we obtain the following result (note how the 0th order is filtered out now):
Unfortunately, this way of handling the 0th order and also the overall quality of the above image leads to the calibration process to fail. Further tweaking of parameters for the blob detector and the different filters etc. is needed. Nonetheless, we almost get all of the dots correctly matched:
When ignoring that the calibration failed, we can run the actual optimization using the camera. The following image shows the simulated result after an iteration:
And here the corresponding captured image:
At the current state, we run into a problem at this stage which will require
some debugging in the future. In the subsequent iterations all the gradients of
the phase mask are None and hence the SGD does not further progress except
reducing the scaling factor of the phase mask which does not change the phase
mask itself. Here the output we get in the command line:
python examples/lens/run_sgd_citl.py
Program mask onto the physical SLM.
15:23:54.833989 Start imshow
Program mask onto the physical SLM.
15:23:58.315224 Start settle
15:23:58.570841 End settle, start capture
15:24:01.862928 End capture
15:24:04.860459 End imshow
15:24:05.442145 Start imshow
Program mask onto the physical SLM.
15:24:08.283236 Start settle
15:24:08.536671 End settle, start capture
15:24:11.972557 End capture
15:24:15.478135 End imshow
- Calibration failed
15:24:16.298924 Start imshow
Program mask onto the physical SLM.
15:24:19.772224 Start settle
15:24:20.023749 End settle, start capture
15:24:20.350141 End capture
15:24:26.692450 End imshow
15:24:26.788523 Start imshow
Program mask onto the physical SLM.
15:24:30.325651 Start settle
15:24:30.580847 End settle, start capture
15:24:30.905616 End capture
15:24:37.320012 End imshow
Traceback (most recent call last):
File "C:\Users\baehl\workspaces\mask-designer\examples\lens\run_sgd_citl.py", line 247, in <module>
main()
File "C:\Users\baehl\anaconda3\envs\mask-designer\lib\site-packages\click\core.py", line 1128, in __call__
return self.main(*args, **kwargs)
File "C:\Users\baehl\anaconda3\envs\mask-designer\lib\site-packages\click\core.py", line 1053, in main
rv = self.invoke(ctx)
File "C:\Users\baehl\anaconda3\envs\mask-designer\lib\site-packages\click\core.py", line 1395, in invoke
return ctx.invoke(self.callback, **ctx.params)
File "C:\Users\baehl\anaconda3\envs\mask-designer\lib\site-packages\click\core.py", line 754, in invoke
return __callback(*args, **kwargs)
File "C:\Users\baehl\workspaces\mask-designer\examples\lens\run_sgd_citl.py", line 214, in main
angles = sgd(target_amp, warm_start_phase).cpu().detach()
File "C:\Users\baehl\anaconda3\envs\mask-designer\lib\site-packages\torch\nn\modules\module.py", line 1110, in _call_impl
return forward_call(*input, **kwargs)
File "C:\Users\baehl\workspaces\mask-designer\mask_designer\neural_holography\modules.py", line 229, in forward
return stochastic_gradient_descent(
File "C:\Users\baehl\workspaces\mask-designer\mask_designer\neural_holography\algorithms.py", line 338, in stochastic_gradient_descent
raise ValueError("Gradient is None!")
ValueError: Gradient is None!
SGD using waveprop¶
python examples/lens/run_sgd_waveprop.py
The simulated results should be similar to the following:
And the corresponding captured image should look something like:
SGD using Neural Holography’s ASM¶
python examples/lens/run_sgd.py
The simulated results should be similar to the following:
And the corresponding captured image should look something like:
Lensless examples¶
Note that for all those examples the scaling issue of the resulting amplitude patterns is very much apparent here. Their size is simply not matching the size of the photo sensor we used. Nonetheless, those images illustrate capabilities of those methods.
GS¶
python examples/lensless/run_gs.py
The simulated results should be similar to the following:
And the corresponding captured image should look something like:
Holoeye¶
python examples/lensless/run_holoeye.py
The simulated results should be similar to the following:
And the corresponding captured image should look something like:
SGD using waveprop¶
python examples/lensless/run_sgd_waveprop.py
The simulated results should be similar to the following:
And the corresponding captured image should look something like:
SGD using Neural Holography’s ASM¶
python examples/lensless/run_sgd.py
The simulated results should be similar to the following:
And the corresponding captured image should look something like:
Camera example¶
python examples/ids_image_capture.py
Those are just two of the images you can expect to obtain when running the script. Several others will be taken that illustrate how you can capture collections of images, rescale/crop them to a given shape etc.
