Is img_as_float more precise than .convert('F') #33

Open
opened 2024-04-03 17:07:23 +02:00 by Benjamin_Loison · 2 comments

I would say so as the result still has the 3 channels for img_as_float contrarily to .convert('F').
Hence computation seems to take more time:

1%|          | 1/100 [04:10<6:53:27, 250.58s/it]

Can proceed in parallel somehow if it is not already the case? It seems to only consume a single core.

Note that I have 16 cores on my laptop and 24 on my computer.

Using gzip only reduces 369M to 343M. Note that if want to mean then have to pay attention as all images may not fit in 32 GB of memory. Could otherwise use an OVH C2-120 with 32vCores and 120 GB of RAM at €1.48 ex. VAT/hour.

Could compute mean in real-time to avoid too much storing.

image

With Pool(30) logs attached.

Should understand what is blocking, otherwise can rent many small instances.

Related to #8 and #19.

I would say so as the result still has the 3 channels for `img_as_float` contrarily to `.convert('F')`. Hence computation seems to take more time: ``` 1%| | 1/100 [04:10<6:53:27, 250.58s/it] ``` Can proceed in parallel somehow if it is not already the case? It seems to only consume a single core. Note that I have 16 cores on my laptop and 24 on my computer. Using gzip only reduces `369M` to `343M`. Note that if want to mean then have to pay attention as all images may not fit in 32 GB of memory. Could otherwise use an OVH C2-120 with 32vCores and 120 GB of RAM at €1.48 ex. VAT/hour. Could compute mean in real-time to avoid too much storing. ![image](/attachments/afbb6be7-ee2d-452c-9e0a-431af861002b) With `Pool(30)` logs attached. Should understand what is blocking, otherwise can rent many small instances. Related to #8 and #19.
543 KiB
6.3 KiB
Benjamin_Loison added the
enhancement
medium priority
quick
labels 2024-04-03 17:07:23 +02:00
Author
Owner

Computing the PRNU estimate on the 100 images flat-field images for NikonD7000

Seems to have possibly interesting PRNU pattern attached (_{0,1} are different GIMP Brightness-Contrast... rendering).

Are we seeing how the camera internally works? What are the yellow dot and circles? Is it related to Wikipedia: Machine Identification Code?

np_array_0_annotated

Is it like figure I a of the internship offer?

image

Screenshot from 2024-04-04 11-54-12

Computing the PRNU estimate on [the 100 images flat-field images for NikonD7000](http://loki.disi.unitn.it/RAISE/Flat-field/Flat-field.zip) Seems to have possibly interesting PRNU pattern attached (`_{0,1}` are different GIMP `Brightness-Contrast...` rendering). Are we seeing how the camera internally works? What are the yellow dot and circles? Is it related to [Wikipedia: Machine Identification Code](https://en.wikipedia.org/wiki/Machine_Identification_Code)? ![np_array_0_annotated](/attachments/8142fb02-ac22-4518-a12c-aaaec6f3e094) Is it like figure I a of the internship offer? ![image](/attachments/1b8a12e7-b960-45a3-b2aa-4e078658195e) ![Screenshot from 2024-04-04 11-54-12](/attachments/69c9d511-1057-427c-b403-92376b043ccb)
Author
Owner

Should use a slider to compare both images like leafmap.org/notebooks/76_image_comparison/.

What would be interesting would be able to do so in GIMP to be able to modify the brightness and contrast.

Related to Benjamin_Loison/gimp/issues/22.

Should use a slider to compare both images like [leafmap.org/notebooks/76_image_comparison/](https://leafmap.org/notebooks/76_image_comparison/). What would be interesting would be able to do so in GIMP to be able to modify the brightness and contrast. Related to [Benjamin_Loison/gimp/issues/22](https://gitlab.gnome.org/Benjamin_Loison/gimp/-/issues/22).
Sign in to join this conversation.
No description provided.