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.


Topics - James

Pages: 1 [2] 3 4
16
Feature Requests / Scaling orthomosaic
« on: March 14, 2018, 11:58:05 AM »
It would be nice, and seems like should be possible, to be able to add scale bars to a chunk after generating the orthomosaic, and the scale of the orthomosaic updating accordingly without having to regenerate it again.

17
Feature Requests / Remove tie points from photos after alignment
« on: December 21, 2017, 12:52:47 PM »
Sometimes I forget to mask parts of an image and photoscan generates tie points there which i would like to delete without having to mask the area and run alignment again.

It would be nice to be able to mask/select the area in photo(s) after alignment and then tell it to ignore/remove these points for subsequent alignment/optimization steps.

I know i can delete the points from the sparse cloud, but when it's just one big mess it's not always clear which points are which, and anyway this would only allow me to remove the valid matches (blue in the photo) and leave the invalid (grey) ones still present which would not be ideal.


18
Feature Requests / preserve current view after operation
« on: December 14, 2017, 07:48:11 PM »
i would love it so much if the 3d view did not reset after the following:
  • align selected photos
  • update
  • optimize cameras
  • anything else that isn't named 'reset view'

19
Bug Reports / Center image on selected marker
« on: April 19, 2017, 07:26:37 PM »
Not sure if this would be a bug report or feature request, but I'm going for bug report.

The old functionality is required when placing markers into sets of images where some are not yet aligned as these would not be included when you filter by marker, and you would still want it to center on the images that are aligned.

Hello Alexey,

Would it be possible to put back something which have been removed since 1.3 update ?

Actually, until 1.2.6, when we were scrolling pictures while a marker was selected in the reference pane, the view was centered on this marker when it was visible on the picture.
Since 1.3, centering on a define marker is only possible right after filtering pictures by marker.

I would like to get back this centering after manual filtering. A workmate of mine, who's working on 1.2.6 can do this... But, working on the lastest update, I can't... :-X

Regards

20
Screenshot #1:
On opening a project in v1.3 (just installed minutes ago after upgrading from 1.2) only a small portion of the model view is displayed/updated. I can rotate the view, but only that small portion in the middle is updated.

Screenshot #2:
Immediately after taking a screenshot, moving the mouse causes the view to be refreshed and all is visible again.

Screenshot #3:
If i don't take a screenshot straight away, and rotate the 3d view, only the central 'viewport' is updated

Screenshot #4:
Immediately after taking a screenshot, moving the mouse causes the view to be refreshed and all is visible again.

21
Python and Java API / Duplicate Chunk - Items to copy, not in api?
« on: January 22, 2017, 06:57:49 PM »
Is there any way to duplicate a chunk, excluding all the bulky things like orthomosaic, model etc, like you can with the GUI command?

chunk.copy() seems not to take any relevant arguments and takes so long on my big chunks it is quicker to do it in the GUI for now.

Using 1.2.6

Thanks

22
General / Align images - Reference preselection
« on: December 20, 2016, 02:41:01 PM »
Hi,

Could anyone shed some more light on reference preselection used for image alignment please?

I have a project with ~26,000 images, all down facing and with xyz camera coordinates that are mostly good but occasionally out by as much as 4 or 5 times the distance interval between consecutive shots (long story). forward and side overlap is just 66% in each direction, which i know isn't ideal but it seems to work well enough for this application.

I understand that reference preselection chooses to compare points between images that should overlap based on camera orientation and position, but does it only choose images where there should definitely be some overlap, or does it over estimate to include images which probably don't overlap but might do if the coordinates weren't particularly accurate? If it doesn't find matching points in images that it thinks should overlap does it then hunt for the next most likely images up to a certain search radius before giving up? I had a hunch it might do that after it aligned some images in a previous project where the coordinates were completely wrong in some parts.

So my main question is how robust is reference preselection when camera coordinates are not so good, and is there a way to quantify how far the algorithm will search to find overlapping images based on available reference data.

Otherwise an 'adaptive' preselection mode which begins by using 'reference' preselection and then uses 'generic' or 'disabled' on images that did not align well might be a good idea.

23
Bug Reports / Chunk Info Alignment Time
« on: September 29, 2016, 05:25:00 PM »
When aligning a subset of images within an already processed chunk the 'alignment time' property in 'chunk info' is overwritten with the time taken to do this most recent alignment, even if it was just a couple of images in a few seconds, so you lose how long it took to originally align the complete dataset, and could be misleading in some cases.

A related question, does the 'matching time' in chunk info also include time to detect points, and if not is this stored anywhere?

Thanks

24
Bug Reports / Vector<T> too long
« on: March 23, 2016, 09:21:12 PM »
error message on pscan 1.2.4 pro win 64

vector<T> too long when building texture using adaptive orthophoto and colour correction with blending disabled.

screenshots attached

error persists with colour correction disabled and mosaic blending mode so seems to be down to adaptive ortho mapping mode.

is it because the model is a smooth cylinder exported from another 3d software?

25
Bug Reports / change path for multiple photos
« on: February 11, 2016, 07:06:59 PM »
am i doing this right?

I select all photos in the photos pane, right click one and select change path, browse for the directory with alternate images (with the same names - is this case sensitive?). a file chooser dialog filter restricts my selection to the image with the same name as the one i right clicked, so i choose it, and then select either 'selected images' or 'all cameras'. in either case only the image i right clicked on has its path changed.

further more, if i right click on an image in portrait orientation then i am told 'image size mismatch' probably due to the way the camera sets the orientation which is lost during processing.

version 1.2.3 build 2331 64 bit

26
General / guided marker placement - blue markers used in optimisation?
« on: February 10, 2016, 07:51:21 PM »
Just wondering if blue markers that are created when placing markers on photos with mesh present (or directly on the mesh) are considered during optimisation, or ignored like grey ones are?

I'm assuming (hoping) only green markers are considered?

27
Bug Reports / align photos - constrain by mask
« on: February 03, 2016, 07:14:31 PM »
In windows version 1.2.3 build 2331 64 bit pro, align features with constrain by mask disabled seems to ignore masked areas of the photos. I thought it should ignore masks if this option was unchecked?


28
General / Build OrthoMosaic Batch Process Projection
« on: February 01, 2016, 11:15:52 PM »
When setting a batch process to build orthomosaics for all my chunks (i have about 25 for this project), it seems to default to a top projection.

I want to be able to do left, right, front and back too.

There is a projection parameter that i can set but this only seems to include geographic projections. There is a 'custom' box but i'm not sure how to create custom projections for left, right, front and back, or if that's the right thing to do anyway.

I shall start working out the python way to do it, and check back here in case there is an easier way before i finish...

Thanks!

**edit** the python was pretty simple and went something like this:

Code: [Select]
import PhotoScan

front = PhotoScan.Matrix([[1, 0, 0, 0],[0, 0, 1, 0],[0, -1, 0, 0],[0, 0, 0, 1]])
back = PhotoScan.Matrix([[-1, -0, -0, 0],[0, 0, 1, 0],[0, 1, 0, 0],[0, 0, 0, 1]])
left = PhotoScan.Matrix([[-0, -1, -0, 0],[0, 0, 1, 0],[-1, 0, 0, 0],[0, 0, 0, 1]])
right = PhotoScan.Matrix([[0, 1, 0, 0],[0, 0, 1, 0],[1, 0, 0, 0],[0, 0, 0, 1]])

doc = PhotoScan.app.document

for chunk in doc.chunks:

if chunk.label.endswith("East"):
proj = right
elif chunk.label.endswith("South"):
proj = front
elif chunk.label.endswith("West"):
proj = left
else:
proj = back

chunk.buildOrthomosaic(surface=PhotoScan.DataSource.ModelData,blending=PhotoScan.BlendingMode.MosaicBlending,color_correction=False,projection=proj)

29
Most of my work is building facades so orthomosaics are generally generated in front projection.

I noticed that with markers turned on in the ortho tab they are projected in top view, even though the orthomosaic was generated from a front projection, so appear in the wrong place.

30
General / Architectural Example
« on: January 20, 2015, 03:04:22 PM »
Hi there,

I don't often (ever?) post results of my work due to the hassle of getting permission to do so, but since this is just a 'by-product' of the actual deliverable, which was simply high resolution 2D orthophotos, i think i can get away with it...

https://sketchfab.com/models/4f8344edaabe47a1bd5ca3e4cb5821ac

James

Pages: 1 [2] 3 4