GPU usage with Hugin

[Álvaro Huertas]

Hi everyone!

I was looking to stitch big datasets of images (300+ images per panorama). I've developed some scripts to automate the process since I have to repeat the same process several times.

I'm following the next workflow:

1. pto_gen

2. pto_mask for each image

3. cpfind

4. autooptimiser

5. cpclean

6. another autooptimiser

7. nona 

8. enblend

I'm finding too slow the process (40+ mins each panorama), specifically the autooptimiser part. I was wondering if there is any way of make this tool to take advantage of the GPU power or maybe antoher kind of optmization in my workflow.

Since i'm new in this kind of pano tools I don't know if i´m following the correct approach but it seems to work fine for my problem.

I can give you more information if needed!

Thanks in advance.

[T. Modes]

alvaroh...@gmail.com schrieb am Freitag, 29. Januar 2021 um 13:38:51 UTC+1:

I'm finding too slow the process (40+ mins each panorama), specifically the autooptimiser part. I was wondering if there is any way of make this tool to take advantage of the GPU power or maybe antoher kind of optmization in my workflow.

Run cpclean before autooptimiser to remove unlikely cp which will confuse the optimiser. (with other words skip step 4).

[Álvaro Huertas]

Then I will not improve the quality of the stitch. I run the cpclean aiming to clean the outliers control points, if I run autooptimiser twice nothing happens and I will get and unacceptable average distance between control points.

[T. Modes]

I said skip the first autooptimiser call (step 4 in your list), not cpclean.

So the sequence is cpfind - cpclean - autooptimiser - …