It seems to me that the time difference between the “Normal” and “High” reconstruction is quite big. So my question would be if it is possible to adjust the settings to get something inbetween. Looking at the “Reconstruction Settings” i see that there are two settings under “Depthmap Calculation”: “Quality level” and “Image downscale”. Is this the only difference or are there other variables in other places? And will i get a result with more points but faster than the “high” setting when i set the quality level of the “high” template to “1”?
I really like the program, but i have some dificulties with separated components in some models. I would like to understand why Photoscan is able to connect the same images? Is the setting in RC more strict while PS is working with weaker points? Will i get the same result when i lower the quality settings (“max features”, “preselector features” and “detector sensitivity” as far as i understand). Or are they using a completely different approach? The camera calculation time is MUCH longer in PS (using “high” in PS).
It would be helpful to have an estimation of the needed disk space at the beginning of a long calculation. And it would be good if you could select multiple temp directories, so that you can use first a SSD and then a hdd or two smaller SSDs.
Hi numerobis0
It seems to me that the time difference between the “Normal” and “High” reconstruction is quite big. So my question would be if it is possible to adjust the settings to get something in between. Looking at the “Reconstruction Settings” i see that there are two settings under “Depthmap Calculation”: “Quality level” and “Image downscale”. Is this the only difference or are there other variables in other places? And will i get a result with more points but faster than the “high” setting when i set the quality level of the “high” template to “1”?
the HIGH quality use 1x downsampled images -> original resolution of images
the NORMAL quality use 2x downsampled images -> 0,5 of original resolution of images
there is no in-between for now…
the quality setting leave unchanged as it is.
I really like the program, but i have some dificulties with separated components in some models. I would like to understand why Photoscan is able to connect the same images? Is the setting in RC more strict while PS is working with weaker points? Will i get the same result when i lower the quality settings (“max features”, “preselector features” and “detector sensitivity” as far as i understand). Or are they using a completely different approach? The camera calculation time is MUCH longer in PS (using “high” in PS).
yes we do it quite different way and we set the alignment a bit strict but it will help you see where are the main issues with alignment. mostly its related to image overlap and image count…
if want PM me some screenshot so can guide you what to do to solve it when capturing images (laser scanns )
It would be helpful to have an estimation of the needed disk space at the beginning of a long calculation. And it would be good if you could select multiple temp directories, so that you can use first a SSD and then a hdd or two smaller SSDs.
This is hard to estimate but say if use NORMAL recon quality the TEMP file will be around 50% of the image data
recommend to clean the cache before running reconstruction. use the WORKFLOW/SETTINGS/CLEAR CACHE
I would like to second the request for multiple drives with a prefered order. I have very nice Optane and SSD drives that I can only take advantage of with small projects, would be nice if I could get a boost from these on the big projects that need the boost the most.
How is the temp space calculated? Is there a 1:1 byte per pixel value we could use to calculate the space on our own? It would be handy as the resolution camera we are using is changing all the time, sometimes on the same project. Is it any different for images with transparency, laserscans or high bit depth?
I did some testing on the demo and seems like RC does not really take advantage the full speed capacity of high end SSDs.
If you look at the task manager, disk utilization would only reach around 10 - 30 MB/s, and that’s observing from photo alignment to mesh reconstruction.
Entry level SSD are rated way more than 30MB/s, so I doubt Samsung Pro or even Optane drives will benefit RC. Its just a waste of money.
I’m actually contemplating if I would upgrade either my current SSD of 850 Evo to 970 Evo Plus or my 2600x cpu. But the way things are looking now, just upgrading the cpu is a wiser choice.