[ptx] auto fine-tune on large images

Rich rich at hq.vsaa.lv
Thu May 19 14:49:08 BST 2005


Emmanuel wrote:
> Le Jeudi 19 Mai 2005 07:39, Rich a écrit :
> 
>>note : i first sent this to pablo privately (didn't check reply
>>address), now i'm adding some more info, too
>>
>>Pablo d'Angelo wrote:
>>
>>>Rich schrieb:
>>>
>>>>auto fine-tune is a nice tool, but on large images autotuning single
>>>>point can take a lot of time.
>>>
>>>How large are large images? And how many of them are in your project.
>>
>>this is a 360 panorama in two rows, totalling 30 images. first i merged
>>both rows separately (by running enblend on them on a separate machine
>>at the end - it had more ram and as a result process was much faster)
> 
> I use to have a desktop computer (AMD2400 RAM:768 Mo, a recent HD 7200tr/mn) 
> from which enblend runned reasonnably fast (as I thought it was already 
> slow). Now that I have a laptop (dothan 1.5 GHz, RAM:512 Mo, "slow" HD 
> 5400tr/mn)., I know what slow is! I have the impression that enblend needs a 
> lot of memory. Does it charges (or need to charge) all images in RAM at the 
> same time??

actually i schedule enblend to run during the night on machines with 
slightly more ram (a couple of gigs). i'm using only hugin for control 
points and initial modifying of pictures on a workstation.

there probably is no way to run engine on one machine and frontend on 
another ? :)

and i tried by image pair on another machine that has slower cpu, but 1g 
of ram - it increases performance a lot. for some points autotuning with 
512mb of ram would take almost 30 minutes, with 1g - something around 
15-30 seconds. so having all your images in ram instead of swap makes a 
huge difference.
-- 
  Rich


More information about the ptX mailing list