Messages

31

General / Re: Camera calibration improving accuracy?

« on: April 28, 2015, 06:49:30 PM »

The process is essentially the same (optimising lens parameters and camera positions) but Agisoft Lens adds in some extra criteria via the target grid that help to (more) correctly arrive at the lens distortion parameters.

Could you explain what these extra criteria are?

Edit: also, how do you know it is more correct, because the Optimize Cameras function in Photoscan doesn't report any results?

32

General / Re: Camera calibration improving accuracy?

« on: April 27, 2015, 08:42:52 PM »

Thanks for the answer, that is good to know!

Am I correct that when I use "Optimize Cameras" that it's basically doing the same calibration as Agisoft Lens? (Since the values are the same).


As an example of how much the real focal length can differ from the EXIF info: when I calculate the focal length for my 50mm lens based on the Agisoft Lens output, I come out at 53,7mm. Quite a difference! (not sure if I did the conversion to mm entirely correct).

33

General / Re: Camera calibration improving accuracy?

« on: April 27, 2015, 07:44:34 PM »

Hello igor73,

PhotoScan uses image EXIF data for the initial step of autocalibration procedure. If the EXIF is missing PhotoScan uses the general guess (50mm focal length in 35mm equivalent), but if this guess is very different from the real values the alignment may fail or be incorrect.

How big is the influence of a small difference between EXIF value and real value? I am asking because lenses often have a slightly different real focal length than what manufacturers state (and list in the EXIF info) For example a 50mm lens might have a real focal length of 48mm or 52mm. Does this tiny difference matter?

34

General / Re: Dense Point Cloud Upper Limit?

« on: April 27, 2015, 11:14:04 AM »

We haven't run into any limits yet, I think our biggest scan was over 2 billion (2000 million) points and I have read forum posts of users who went even higher.

If you process such a big scan in one go, then memory can be an issue even for the DC stage (normally the DC stage doesn't need that much memory). This also depends on the nature of the scan, a scan of a terrain (photos in a grid pattern) will take less memory than a complex 3D object.

My advice would be to process the Dens Cloud at Medium first and see if that gives you the detail you need. Because if you process it at High or Ultra and you end up with a point cloud that is un-workable due to the size, then you have wasted many hours of processing (since there is no way to reduce the point cloud at this point).

If do you find a program that can mesh such very dense point clouds then please let us know. 

35

General / Re: dense Clod from 1 or 2 Pic

« on: April 24, 2015, 05:39:43 PM »

Photoscan cannot extract depth information from a single picture, you need 2 pictures at the very minimum (but more pictures is better).

If your pictures do not align, this is probably because the quality of the photos is not good enough (or your subject does not have enough details, you cannot scan featureless or shiny objects).

Show us some example photos and it will be easier for us to help you (full resolution photos would be best).

36

Feature Requests / Re: high frequency noise reduction

« on: April 24, 2015, 02:44:19 PM »

Noise reduction / point cloud filtering is one of our biggest requests, would be very useful to have.

37

General / Re: Camera calibration improving accuracy?

« on: April 24, 2015, 01:11:41 PM »

We are never using camera calibration, and are getting excellent results. However, we are using really good lenses and correct for lens defects (distortion, chromatic abberation, vignetting) when converting the RAW files.

I can imagine that camera calibration is useful when using (for example) fisheye lenses or other lenses with a lot of distortion (some zoom lenses can also have very weird distortion patterns).

38

General / Re: estimate image quality

« on: April 23, 2015, 11:32:37 PM »

It has no unit, it goes from 0 to 1.0 (where 0 is bad and 1 is good).

39

General / Re: NVIDIA Tesla Benchmarks

« on: April 20, 2015, 08:57:32 PM »

Are you using server fans? Because these fans are small but displace a lot of air (and are LOUD, another reason not to use this case XD)

That's strange, because that is how these kind of cards are often used:

http://www.beyond3d.com/images/articles/tesla/S870_34_B-big.jpg

http://www.nvidia.com/docs/IO/43395/NV_DS_Tesla_S1070_US_Jun08_NV_LR_Final.pdf

You can see the row of fans in the front. This is an nVidia Tesla S1070 case, you can pick them up for cheap on eBay. The included M1060 GPUs are not  good for Photoscan (around 350 million samples/sec or something).

You can't put consumer cards in it because of the mounting pins (and using 4 cards is not very effective anyway). Not sure if your Tesla cards would fit either since they are newer models.

40

General / Re: 16GB or 24 GB RAM for Photoscan

« on: April 17, 2015, 04:21:13 PM »

As far as I know, there currently are no i7s that are capable of using more than 64GB RAM:
i7-5960x (Intel Spec Page)
That being said, if he's buying a Dell through his university (like we did) he'll have no choice of motherboard. Also, I'm pretty sure a Xeon can fit into that MSI motherboard.

Ah, you are right, it does need a Xeon for 128GB. But at least you do not need the more expensive dual CPU Xeons.

41

General / Re: 16GB or 24 GB RAM for Photoscan

« on: April 17, 2015, 12:38:53 PM »

you can go with a super high quality gaming PC (I recall the i7-5960X is one of the best these days) but you won't be able to use more than 64 GB RAM.

There are single socket (non-Xeon) motherboards that support 128GB RAM with 8 memory slots (using 16GB sticks keeps it more affordable than having to use 32GB sticks) , for example:

http://hothardware.com/news/Holy-Overkill-MSIs-X79AGD45-Motherboard-Supports-128GB-of-RAM

42

General / Re: Titan X benchmark

« on: April 13, 2015, 10:08:52 PM »

I agree with Wishgranter, it's not a PCI bandwidth issue.

If you look at the GPU load graph during processing, you can see that the work is done in batches. A GPU gets a batch of work and has high load for a minute or two, and then goes idle for a short while until it gets a new batch. At the end of the Dense Cloud reconstruction one GPU is often idle while the other is finishing the last batch.

So with a single GPU you have often 100% utilization, and with two GPUs it's just less efficient.

43

General / Re: Agisoft Vs Competitors

« on: April 13, 2015, 10:05:47 AM »

1 - Why (what is the reason for that)?

No specific reason other than Autodesk having a habit of buying every 3D program out there (and development of these programs don't necessarily benefit from getting bought by Autodesk).

-a very customised work workflow.  Once you understand the setting and parameters you can tune your workflow for best or fastest results very easily.  The parameters can also be changed half way through, unlike Pix4D

I think if Agisoft would spend some time on updating the documentation and making the programs settings a bit more clear, this would make Photoscan much simpler for beginners.  There is a lot of confusion among beginning users because none of the settings are explained in the program itself (and compared to other programs the settings are often not self-explanatory). As a new user you have to search the documentation (and often the forum) to find out what things do.

If every setting had a small tooltip or a link to the documentation explaining the setting then this would make a big difference in usability and make the program less scary for new users.

44

General / Re: Agisoft Vs Competitors

« on: April 11, 2015, 03:25:47 PM »

Hi,

Why Autodesk has so many products

I'm just happy they did not purchase Agisoft/Photoscan already.

45

General / Re: 16GB or 24 GB RAM for Photoscan

« on: April 10, 2015, 10:30:14 AM »

Put in as much RAM as you can afford / fits. For larger scans the meshing stage will need a ton of memory, and even 24GB will probably not be enough.