Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Topics - maddin

Pages: [1] 2 3
Feature Requests / Gradual select By Angle for polygons
« on: March 27, 2017, 12:26:40 PM »
In order to clean up a dense reconstruction, a 'Gradual Select > By Angle' would be great.

So we would select a bunch of polygons, then run this new tool and get a slider to set a crease angle.
The polygon selection would be grown until the algorithm encounters a polygon whose angle to the already selected polygons is greater than the selected crease angle.

This tool should also run in 'deselect' mode, such that we can region-select a larger group of polygons, and then use the tool to deselect planar parts. All the non-planar parts could come from noisy or interpolated data, and this tool would allow for easy selection and deletion of such parts.

Feature Requests / Some basic 3D modeling tools with camera projection
« on: March 27, 2017, 12:07:23 PM »
I constantly find myself wishing for some simple 3D modeling tools inside of PhotoScan, in order to fix up some glitches in the dense reconstruction.

While I can easily export the mesh and edit it in some 3rd party application and bring that back into PS again, it would be MUCH better and faster if I could do the modeling while the camera projection is active inside of PS, so I can precisely see where to move my geometry such that it nicely aligns with the color information.

Some suggestions for such basic tools: a Cut tool, an Extrude tool, a Move/Rotate/Scale tool for vertices/edges/faces

Feature Requests / Volume carving tools
« on: March 27, 2017, 12:02:35 PM »
It would be really helpful if PS had some tools to draw outlines in individual photos and then use those outlines to do some sort of volume/space carving (by simply intersecting all the cone-like 3D structures that you get by extruding each outline from the camera position).

This would allow to quickly reconstruct some user-defined volumetric shapes that could optionally be refined by dense matching?

Feature Requests / Tools for wall surfaces
« on: March 27, 2017, 11:58:49 AM »
PS is great for organic objects, but I am having a very hard time getting good results for walls. Interstingly, walls would be really simple to model, but the workflow between PS and a 3rd party 3D app is pretty cumbersome.

It would be great if I could help the dense reconstruction in PS by indicating regions that contain a smooth, even, uniformly colored wall (maybe by picking two parallel lines on the wall?).

With such a tool, we could help the reconstruction for cases where there is little pixel information, but easy large-scale information...

General / Crash to desktop while reconstructing mesh
« on: March 01, 2017, 01:26:31 PM »
I have a medium-sized project that crashes PhotoScan Pro 64bit 1.2.6 AND 1.3.0 to desktop when reconstructing the mesh from the dense point cloud.

Attached is the details page of the report in 1.2.6, using a mesh based on the low-res point cloud.

Any ideas what could be wrong?


Feature Requests / Batch processing: Saving multiple versions
« on: February 28, 2017, 04:01:39 PM »
It would be great if we could use Batch processing to evaluate multiple settings and save those results independently; this way, we could use a weekend to run several versions to learn which settings work best.

I realize we can export at various points in the pipeline but hat does not contain all the additional information, error metrics etc.

Feature Requests / Batch processing: Python
« on: February 28, 2017, 04:00:17 PM »
Would be great to be able to use Python code as part of the Batch Process pipeline, e.g. by inserting a "Run Python code" element into the batch queue?

General / How to keep only the best data?
« on: February 27, 2017, 04:10:17 PM »
What is the recommended strategy to keep only the very best data in an aligned project?

Let's say I have a project with many cameras but the overall result (dense cloud, mesh) isn't great. So I would like to have some systematic way of pruning photos/cameras/points/markers from the project that are less reliable than others. The result would probably something with less coverage but hopefully those parts covered are really nicely reconstructed.

To start with, we have the 'Image Quality' that somehow measures the sharpness of the sharpest part of the image. Is that a good measure and what threshold should I use?

I also know about the Gradual Selection tool but those unitless thresholds seem very arbitrary, too, and don't tell the complete picture each on it's own (e.g. I could throw away all points above a certain reprojection error, but then maybe the angle between the cameras isn't large enough so their 3D location is rather weakly defined (this is covered by Reconstruction uncertainty IIRC)), so I am looking for help how to combine these. Also, Gradual Selection does not apply to the dense point cloud, does it?

Then there are some reprojection numbers in the Chunk Info, as well as some error numbers for each camera in the Reference Tab ('Error (pix)'). 

So now the question is how to combine these all into one consistent approach to only keep data that looks very reliable?

Thank you for any insights

General / Look Through - should the view match the photo?
« on: February 24, 2017, 07:37:18 PM »
After aligning and processing a project, we can us the 'Look Through' function on photos to set the perspective view to that of the camera.

In my case, that view does not really match. The general direction in view content is somewhat ok, but the field of view, for example, is completely off. My images were taken with a 16mm lens on a full-frame DSLR, which results in a hFOV of about 97 degrees, but the perspective view uses the default 30 degrees.

Shouldn't the photo and the model view match closely when 'looking through' a well aligned camera?

Feature Requests / Better names for workflow parameters
« on: February 23, 2017, 08:27:15 PM »
Some of the parameters used in a normal reconstruction workflow are not very useful IMHO.

1.) Pair preselection - Why call it "Generic" when it really tries a low-res match first?

2.) Alignment Accuracy - "High" really means "use full original image size" - why not use "Full Size", then? Also, calling this 'Accuracy' in general is somewhat misleading IMO. Something like "Image Resolution" would indicate better what it really is about.

3.) Along those lines: The current names we have for the dense point could quality level ("Ultra High", "High" etc.) are not terribly informative, too.  At the very least, they should be augmented with their actual meaning, i.e. "Ultra High (full picture size)", "High (half size)" etc.  And it is somewhat confusing that "High" in the Alignment is the same as "Ultra High" in the dense point cloud reconstruction.

That way, users can an idea of what really is happening without having to look it up in the online help.

Feature Requests / Show photo zoom factor
« on: February 23, 2017, 07:46:57 PM »
When zooming into a photo in 2D view, it would be helpful to display the current zoom factor, similar to how the field of view is shown in Perspective 3D view.

General / Can alignment be seeded/improved by markers?
« on: February 22, 2017, 09:18:57 PM »
Is it possible to help alignment by providing a bunch of markers for each image?
Will this improve initial alignment or only optimize the alignment afterwards?


Bug Reports / Problem with using panoramic image for texture building
« on: February 10, 2017, 02:06:14 PM »
So I have this project of about 300 aligned DSLR images, along with a few full-spherical panoramic images, also successfully aligned. Also, I successfully computed a dense point cloud on 'High' and a mesh with about 2.9 Mio. faces.

Next, I do a Build Texture > Single Camera - but PS just sits there, trying to generate the texture atlas needed for this single (panoramic) image and never returns on me. I left it sitting at 0% progress bar for 45 min. but no apparent progress was made.

Note that I can use 'Keep UV' and disable all cameras but the one panoramic image and this works fine. So it seems not to be the actual projection of the image data but rather the generation of the UV coordinates from a single panoramic camera that is the problem.

Anyone else encountered this before?

Feature Requests / Trigger PS actions from another application
« on: February 07, 2017, 05:08:27 PM »
It would be very useful if we could (using COM/OLE, for example) trigger certain calculations in an already running PS session.

For example, when creating a model in an external 3D application, the steps to import the model, calculate and export the new texture map could be triggered from outside of PS.

Python and Java API / Measuring contribution of camera?
« on: February 02, 2017, 08:57:41 PM »
Say I have a project with many overlapping cameras, all aligned, model built.

Is there a way to measure how 'important' a camera is to the final texture build?
In other words: Is there a way to determine how many pixels of the final texture came from each camera?

One idea would be to swap each camera image with a unique solid color, then compute the texture (without any blending?) and then count the occurrence of each color in the final texture image. But how to do that in practice inside of PS Pro 1.2.6?

I am asking because I try to find out which of my many cameras I could discard for texture building, in order to avoid blending artifacts.

Thanks for any suggestions

Pages: [1] 2 3