Agisoft PhotoScan 1.3.0 pre-release

[venkatg]

Hello Alexey,

I've been using v 1.3 for a few days now and I'm loving the command line feature!

Are there any plans to introduce Photoscan as a dynamic library (.so) so we can seamlessly use import and use the Photoscan module with an IDE? It would greatly improve the python scripting experience, particularly for users interested only in headless mode.

Cheers,
Venkat

[KellyNess]

Locks up upon opening a pretty large .psx file - 2000 photos... File worked fine with 1.2  anyone else have a problem?

[mrc]

I'm having an issue with markers.

I am scanning small objects and include a scale bar positioned near the object in some photos. My final model never includes these scale bars, I crop them out in the alignment stage as they often cause noise.

In previous versions, my workflow was to finish up to the model stage, then place markers on the scale bar in the first photo, find a second photo with the scale bar and place the markers again (the line comes up to show the 2d line detected and I find the intersect on that line and set the marker). I then create a scale bar and give it a value.

In this version, even though my scale bar is not within my model, the system acts as if it knows the 3D point and places the marker directly from the first photo. i don't have to find the intersect on a second photo. The only problem is that this 3D point is wrong, and even if I drag the marker to the right point in the other photos, my scale is way out in the final model.

Is there a way to return to the old behaviour?

Thanks
Michael

[mrc]

Actually, further to above, it seems like some of the tie points I am deleting are not actually getting deleted. I generally mask everything other than my main subject, but the last few objects I have processed with this new version have reconstructed objects in the masked areas, even though I deleted tie points in those masked areas in the sparse cloud before dense reconstruction. Definitely something odd going on!

Edit--Actually not sure what caused above as regenerated and the other points didn't show up. Will see what happens on the next scan.

[Alexey Pasumansky]

Are there any plans to introduce Photoscan as a dynamic library (.so) so we can seamlessly use import and use the Photoscan module with an IDE? It would greatly improve the python scripting experience, particularly for users interested only in headless mode.

Hello Venkat,

At the moment we are not planning to supply an external library or SDK that allows to access PhotoScan algorithms.

[Alexey Pasumansky]

I'm having an issue with markers.

Hello Michael,

If you need to reproduce the marker placement behavior with the equipolar lines, you need to do the following:
- Add new marker in the Workspace pane by the corresponding button,
- Open image in the Photo view and put marker projection using Place Marker option in the context menu instead of using Create Marker tool,
- After that you'll have the marker with the single projection only and will see the equipolar lines on another images.

In the version 1.3 PhotoScan is always using Guided Marker placement approach, even if only sparse cloud is present (in the previous version it worked only starting from the dense cloud).

[Alexey Pasumansky]

I don't know if this has been proposed, but I think it would be helpful to be able to create multiple point clouds within the same chunk. That would help in cases where you would like to reproduce certain areas of a project in high accuracy while others in lower accuracies (ie ground data)

Hello cadm8,

It's better to put the suggest to the Feature Requests forum section:
http://www.agisoft.com/forum/index.php?board=11.0

At the moment we are not planning to implement the support for multiple instances of the sparse cloud, dense cloud, mesh and etc. for the same chunk, as it may lead to the confusion.

[Alexey Pasumansky]

Hi Alexey, unrelated to my recent bug post for 1.2.6, I noticed that 1.3.0 appears to have an issue with processing photos at the edge of chunks.  See this image:

In version 1.2.6, there is not a gap between the upper two chunks when the data is processed. In 1.3.0 data is lost at the edges.

Hello Mark,

Could you provide some additional details to support@agisoft.com regarding the issue. Probably the screenshots from both versions, so that the difference can be clearly observed.

[MManning]

Hello
with the command line feature of v1.3 it would be nice to submit the PS model build to a GPU node on a cluster. the issue is that when you submit a job to a cluster you dont know which GPU node will be assigned the job. Since I cant install the license on every GPU node, I get a license error as the result of my run. ( the PS install itself is in a container which is executed on the node). Does anyone know of options to allow for this type of processing?

Thx!

[Mohammed]

Dear Alexey,

-I tried to update the new version in my laptop but when i put the license the software give me that '' Activation Key is already in use'' what dose that mean? in previous version it works fine on my PC and laptop now i can only open it on PC.

So is that mean the new floating license work on one computer not PC and Laptop?

Mohammed,

[Jan Philip Göpfert]

After importing an existing camera calibration via the GUI, PhotoScan automatically initiates building of the coarse point cloud. The Python API does not seem to offer this possibility (or does it?), and so, after importing via chunk.importCameras(camera_path), no dense cloud can be constructed. How would you work this using Python?

On a related note -- can you reset the calibration of specific cameras and then recalibrate them via chunk.alignCameras(recal_cams)?

[Alexey Pasumansky]

Hello Jan Philip Göpfert,

After importing cameras via Python you need to use buildPoints() function. Using alignCameras() will discard imported camera positions and will estimate them from scratch. Actually, it's not a version 1.3 feature, it was present in the older versions as well.

[Alexey Pasumansky]

Hello tornado2011,

I think that you have a node-locked license (since floating license policy has been only introduced recently with the new activation system), so one license key can be only used on a single machine. If you need to transfer the license, you need to deactivate the license first.

[bisenberger]

In the Model view front faces are transparent and back faces are visible in Shaded, Solid, Wireframe, and Textured modes.

[Alexey Pasumansky]

Hello bisenberger,

Do you observe it on any mesh generated in 1.3?