Forum

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.


Messages - James

Pages: 1 ... 44 45 [46] 47 48 ... 50
676
General / Re: interesant application - removing SHADOWS
« on: June 17, 2013, 02:34:11 PM »
If photos/model are georeferenced and timestamped then photoscan could calculate the position of the sun automatically without needing to be too clever.

edit: It would only help with shadows cast by the object itself though, not things out of camera view like trees etc.

677
General / Re: Strange workstations testing results.
« on: June 15, 2013, 07:09:21 PM »
Don't suppose it makes a massive difference but did you try parallel processing identical chunks with the same images, or completely different chunks with different images? Just wondered if there may be any caching of anything anywhere in that case that might make the results look better than they really are. I doubt it really but it might have an effect.

Brilliant work though :)

678
I can recreate this behaviour exactly by feeding in a non-existent file name to doc.open(). (It works perfectly if i put in a valid path)

Make sure your path names are being constructed correctly, maybe you need to add another '\\' between the directory and the project file name?

The 'NoneType' object has no attribute 'matchPhotos' error means that you have a null object, in my case caused by attempting to open a file with an incorrect path, so nothing to do with parameter specification.

679
Feature Requests / Re: Multiple Masks Per Photo
« on: June 14, 2013, 11:48:43 AM »
It had occurred to me also that this could be useful sometimes.

Think I had probably aligned some photos with mask constraint, and then modified the masks for the purposes of texture generation, then realised i needed to align the photos again for some reason.

I guess the export/import mask options would allow this to be achieved? I haven't used them much i must admit.

680
General / Re: need help here
« on: June 12, 2013, 12:51:22 PM »
I think your photo IMG_4623.JPG is the weak link. The 2 photos before it align, and the 2 photos after it align, but IMG_4623.JPG does not have sufficient overlap with either the first or last pair to join the sets together.

The is a 'horizontal' strip of land running through the approx middle of IMG_4623.JPG that does not appear in preceding or following photos so there is no way to connect them.

681
Bug Reports / Re: Single camera UV mapping and occlusion
« on: June 11, 2013, 10:22:01 PM »
This also seems to apply to faces that are only partially within the field of view of the camera i.e. on the edge of the photo, and is more of a problem (to me) the larger the faces are, i.e. after decimation.

682
Bug Reports / Single camera UV mapping and occlusion
« on: June 11, 2013, 07:42:32 PM »
It seems that when you select single camera mapping mode, it is sometimes hit and miss whether faces are UV mapped if they are partially occluded from the selected camera position.

Image below shows 2 partially occluded faces (left) are UV mapped (right) but the face between them has not been.

683
+1

drag and drop please!

684
Feature Requests / Filter photos by enabled/disabled status
« on: June 10, 2013, 04:45:24 PM »
I'd love to be able to hide photos once I have disabled them. I seem to be disabling ~90% of my photos before building texture most of the time.

Possibly even better would be to dim/grey out disabled photos in the photos pane but leave visible.

It is hard to tell at a glance which photos are disabled just by the little red disabled icon, especially when markers are selected and you also have the marker icon on each photo, disabled or otherwise.

685
Feature Requests / Visual display of 'per image' marker error
« on: June 10, 2013, 03:48:58 PM »
When you position a marker it turns green and photoscan's inferred position (grey flag) disappears. I would like an option to leave it visible or some other way of showing the disparity i.e. error bars, numeric x,y pixel error etc.

If I am right in thinking (in my ever simplistic way) that the error values shown in ground control are based on the difference between where the markers have been placed and where photoscan thinks they should be, then it would be really useful to know where it thinks they should be.

Sometimes I stare at a project far too long trying to work out where an error is originating from and eventually realise it is because I put a marker in a stupid place. Most of the time photoscan's estimated positions are more accurate than me!

Maybe this is because I vary rarely work with printed targets and more often building features that are surveyed using scanning and total station so is too easy to put markers in the wrong place...

686
Camera Calibration / Re: Undistort Photos
« on: May 31, 2013, 03:58:25 PM »
That's fantastic I shall give it a go, thanks!

687
General / Re: New to ortophoto
« on: May 30, 2013, 01:24:07 PM »
Looks like many of the cameras are resolved in pairs to the same position. If this is because they were actually taken from the same position (or very close) I think you will get much cleaner results by removing these 'duplicates'. The short baseline distance is no good for determining depth, but will generate a lot of good matches as the images are so similar. I'm not sure how well photoscan filters out photos taken in the same position like that, if at all.

Really nice to see the survey in action though, and not just the results :)

688
If you build geometry and cancel before it completes creating depth maps, then returning to the build geometry dialog allows you to reuse those depth maps even though they are incomplete, giving you an incomplete or strangely formed model

If the previous step had not even finished loading photos then you can still 'reuse' depth maps that had not even begun being created and no model is made.

Perhaps this is not a bug and maybe even a desired behaviour or undocumented feature :)

689
General / Re: No depth maps for some camera positions
« on: May 29, 2013, 10:35:29 PM »
Could it be a case of insufficient overlap and/or too much difference in camera orientation.

I.e. photoscan detects hundreds of 'candidate' matches, but can not calculate any transform that results in a good pair of camera positions, because the candidate matches are basically wild shots in the dark and the pair of photos are too just different.

When you look at the matches between a pair of aligned photos you can see there is generally a strong correlation between matches connected with blue lines (valid) and no correlation between matches connected with red lines (invalid).

I don't know the algorithm for sure but I guess that there is an initial pass to detect features at a small scale, then match them between pairs of images without reference to the bigger picture, and finally an iterative process to identify the matches that make sense in terms of a basic camera translation/rotation etc.

If the red lines between your invalid matches look like a caffeine driven attempt at a spider's web then the problem is probably with the photos rather than the software.

690
General / Re: Unlimited technology - Large GIS presentation stuff
« on: May 28, 2013, 04:26:52 PM »
Just found it.

Consider me convinced. Very impressive!

Pages: 1 ... 44 45 [46] 47 48 ... 50