Cannot change exposure and frametime on Aurora A anymore

Hello,

I am currently re-running some code which worked on LightOn Cloud in Sep / Oct 2020. In my code I set the OPU exposure and frametime as follows:

from lightonml.projections.sklearn import OPUMap
mapping = OPUMap(n_components=num_rand_proj, verbose_level=1)
mapping.opu.device.exposure_us = 600
mapping.opu.device.frametime_us = 1000
y = mapping.transform(opu_input.astype(‘uint8’))
mapping.opu.close()

where num_rand_proj is the number of projections and opu_input is the data I want to project.

When I use Aurora A the system prints:

OPUMap was not fit to data. Performing fit on the input with default parameters…
OPU: random projections of an array of size (64975,)
OPU: using frametime 500 μs, exposure 400 μs, output ROI ((0, 512), (2040, 64))

which indicates that I have not been able to successfully set the exposure and frametime to 600 and 1000. Aurora B and C are not available for booking so I could not test this there.

Please can someone help me with this?

1 Like

Hi Sidharth,

unfortunately Aurora B and C are undergoing a maintenance procedure. We will look into the exposure and frame time issue and come back to you as soon as possible!

Charidimos

Thanks for the quick response!

Hi Sidharth,

Indeed exposure and frametime being features for advanced users, we haven’t advertised changes on this. Since version 1.2 of lightonopu, the OPU must be fitted before transform, and exposure and frametime can be specified in the **override args:

mapping = OPUMap(n_components=num_rand_proj, verbose_level=1)
mapping.opu.fit1d(opu_input.astype('uint8'), exposure_us=600, frametime_us=1000)
mapping.opu.transform(opu_input.astype('uint8'))

(be sure to call the transform method on mapping.opu instead of mapping, or you’ll have a conflict.
See also the full release notes for more information.

Hope this helps,

Charles

2 Likes