Forum

Author Topic: Agisoft Metashape 2.0.0 pre-release  (Read 63517 times)

flyzk

  • Jr. Member
  • **
  • Posts: 71
    • View Profile
Re: Agisoft Metashape 2.0.0 pre-release
« Reply #30 on: October 22, 2022, 01:25:52 PM »
Hello flyzk,

If you don't know yet which version you are going to use it is better to start with Metashape Professional 1.8 and activate the license there. Such activation will work for both Metashape 2.0 and earlier versions.


Yes but, is there any risky that by activating licence with 1.8 that I should paid for a future upgrade sooner that by activating licence in 2.0 , in case of your business model would change ?
I mean, is there any negativ point by activating in 1.8 or potential negativ point in the future?

thanks.

vmpakas@culture.gr

  • Newbie
  • *
  • Posts: 4
    • View Profile
Re: Agisoft Metashape 2.0.0 pre-release
« Reply #31 on: October 24, 2022, 11:53:44 AM »
dear Elizaveta


i sent you the print screen of the workspace pane you requested. It is obvious that imported point cloud is used  as laser scans
The last thing i  tried is to reset  the Laser scan alignment. This way i finally have a relative solution between images and laser scan. Note that i  have only one laser scan in this test project.

best regards
Vasilis


e.spiridonova

  • Agisoft Technical Support
  • Newbie
  • *****
  • Posts: 38
    • View Profile
Re: Agisoft Metashape 2.0.0 pre-release
« Reply #32 on: October 27, 2022, 01:15:58 PM »
Yes but, is there any risky that by activating licence with 1.8 that I should paid for a future upgrade sooner that by activating licence in 2.0 , in case of your business model would change ?
I mean, is there any negativ point by activating in 1.8 or potential negativ point in the future?
thanks.

Dear flyzk,
All the users that have previous versions of Metashape can download to 2.0 for free. 2.0 version comes with the new activation system which means that new activations made with Metashape Professional 2.0 will use a new license format incompatible with Metashape Professional 1.x versions. Update to 2.0 does not require any additional actions from user - only to install the version and run it with the previously activated license.
Best regards,
Elizaveta Spiridonova,
Agisoft LLC

wojtek

  • Sr. Member
  • ****
  • Posts: 284
    • View Profile
Re: Agisoft Metashape 2.0.0 pre-release
« Reply #33 on: October 28, 2022, 12:18:32 PM »
Hey Agisoft Team,

Are there any improvements in regards to the reconstruction algorithm in version 2.0.0 or is it largely the same as 1.8.4?

I am wondering if I should test it to see if there are less weird errors going on in the new version.

Paulov

  • Newbie
  • *
  • Posts: 44
    • View Profile
Re: Agisoft Metashape 2.0.0 pre-release
« Reply #34 on: October 28, 2022, 08:12:01 PM »

thanks to that last response I got the markers working on a point cloud, I was getting nuts.

I was testing the beta trying to trying to follow this tutorial

https://agisoft.freshdesk.com/support/solutions/articles/31000151726-how-to-import-external-point-cloud-e-g-lidar-to-metashape

I could not add the markers, but thanks to this final answer I could.

Please add to the tutorial the  "Use as laser Scan" information in order to add the markers to the cloud, in bold letters :D


I'had already aligned independently the images, wonder if having the markers now added  it will allow me to merge chunks or i'll need to re align...

carligon2012

  • Newbie
  • *
  • Posts: 4
    • View Profile
Re: Agisoft Metashape 2.0.0 pre-release
« Reply #35 on: October 29, 2022, 07:54:11 AM »
I wanted to know if you have been able to solve the problem that processors with many cores have, in the tests carried out by Pubget Systems https://www.pugetsystems.com/labs/articles/Metashape-Pro-1-7-2-CPU-Performance-Roundup-2261/  it is clear that a ryzen 7 5800x (8 cores) has more performance than a ryzen 9 5950x (16 cores) which it doesn't make any kind of sense, from what I saw the problem was seen in the tile model but there have also been problems with the dense point cloud, it is assumed that the 5950x will tend to take less than twice as long as the 5800x or something approximate but no is the case, even the ryzen threadripper 3975wx(32 cores) is slower than the 5800x. I have done my own tests but only on my laptop that barely has 4 cores, so I could not draw any conclusions about it, it would be very helpful if you can answer that question, and the last question would be, how recommendable is it to use memories DDR5 in metashape, are they worth it?


Paulov

  • Newbie
  • *
  • Posts: 44
    • View Profile
Re: Agisoft Metashape 2.0.0 pre-release
« Reply #36 on: November 01, 2022, 08:15:08 PM »
I tested the beta in regards to the LIDAR meshing and aligning with photos, photos without any GPS or positioanl data.

Initially I followed the tutorial but the tutorial uses prepositioned cameras with GPS so the workflow changes a bit.

NOTES:
-many times when placing markers on LIDAR it is created in the correct position but in Reference panel XYZ info is missing. Sometimes I need to add a marker 3-4 times for XYZ values to show up. ACTUALLY, afer messing around what occurs is that if right click placemerker-->new marker that allways fail to return XYZ but if right click-->Add marker, that seems to work, but in that case the marker enable check is off so is not visible until you go back to Reference panel to check it on.
 
-once  marker placed on LIDAR seems not possible to fine tune position, unlike it is possible in 2d photos.

-when moving marker for fine tune in photo, please add a zoom rectangle in the window corner to pin pint it.

-after cameras are aligned with the laser scan, if I try to finetune adding marker in the scan and then adjust in the image, after moving the blue flag on the image the point in 3D scan gets moved! WTH! Shoulntd alsoshow up as green flag on the img? 

-ONCE, when you save the project, the markers on the LIDAR data where lost !! [This only happened the first time], also the cloud moved to somewhere else!!. When this happened, I saved overwriting a previous attemp project, no sure if that oculd be related. Second and thir time I saved new project all worked fine. fingers crossed.

-when importing LIDAR if you select WGS84 EPSG 4326, after import, when you try to center the view, you get "Unsuported Datum" error window.
If you double click on the LIDAR in the Workspace, being it in an scan fixed group, app crashes


-tutorial fails to explain that in order for the markers to propagate properly, need to be matched in at least 2 photos. This is because tutorial aerial photos already are prepositioned.

-at some point my point cloud showed as NA and I had to reimport due to missing alignment, not sure what I did to ge this to happen.

-Overall the normal computation is accurate for a complex dense point cloud, yet ocasionally some areas like traffic lights or signs have the normals inverted, a tool to refine / armonize the normals of the point clouds would be nice. The meshing of dense point clouds is one of the best out there. I may revisit Meshlab since I've not tested for long but is the best meshing I recall to have seen.

-3d meshing is not possible directly from scan, it must be set as dense poitn cloud otherwise it dos not appear as source option in the dropdown when generating mesh.

-3d mesh is overwritten unless it is duplicated, changing the name won´t help, at least a warning of rewrite would have been nice.

-Intensity of the LIDAR seems not to load, cloud is black (I'm using .las), there should be some visulization options like the saturation of the sclarfile, like in Cloud Compare. I had to "boost" the intensity in Cloud Compare to check if that would make it visible in Metashape but still black. Aparently there is no way to "refresh" the point cloud in Metashape after saving it in Cloud Compare so I had to remove and reimport, yet, still black, I think this is a bug.

-LIDAR available scarlar fields are predefined, should instead show the ones in the .las file, like the missing scalar field GPS time. Or add a "custom" button and then read the ones in file and let user to choose.

-May be there is already option, I miss a quick way to change point size in runtime like press control and mouse wheel up down.


I miss the hability to have a temporal clipping volume, to isolate an area of the cloud in order to place markers, as it may be partially obstructed by other elements. Or even siimply to work ate higher fps rate.

I was unable to perform an incremental alignment, after importing new cameras, adding markers, being one part alread aligned it was not possile to align extra photos. warning meesage was shown saying new  photos would not be ligned, I should check manual for incremental alignment, I did and all was set as should in theory. I had to re-run the whole alignment for all the photos by reseting previous.
I found this strange to happen, since I was aligning a sequential iamge set, and for the first part, when it failed I was able to select the failed cams, reset alignment, re run alignment for those cameras and the end ressult was really good. I can´t see why adding theese new, even with markers that bind the new cameras to the laser scan present, not possible to align, may be the markers need to be shared between images?
The second part does align in a separate chunk. May be right now, using sequential the only way is to compute inseparate chunks and later merge chunks. Since the reference LIDAR cloud is hte same, the markers will be valid on both sides, I guess.
The scan is a 2 level road so from one camera set is not visible the other. I used sequential at first but even with stimated it did not work until

a potentially glitchy behaviour is tat ocasionally the scans will stop being rendered, this seems to happen in duplicated chunks cointaining scans.


-I understand LIDAR will be tied to the pro license, but between pro and STD hte price gap is massive, may be at some point LIDAR module ass a DLC for the base would be superb, this is becasue nowadays there are a lot of free public lidar souces.

even if chunks are dissabled, they keep bein visible in model space should this happen? it's a mess.

It is happening to me that in the copied  chunk the scan, being set a scan data is not present as option as source mesh, even if I unset as scan data and is set as dense poitn cloud, nothing is dissabled.

I wish it was possible to move stuff from one chunk to another easily by dragging, like the 3d model. If I accidentally missaling the 3d model on the copied chunk I would like to simpy do a quick transfer from the original chunk. It just happened to me that after adding markers to the second set of cameras, the meshed point cloud scale has gone nuts. not sure if a bug or expected behaviour. I see there is no lock transform option for models.

Merging chunks trashes the 3d model scale/position. Also scan clouds are not merged so I needed to reimport and remesh the scan.


MISC
camera gizmo plane scales over distance, not sure if I like this...
camera gizmo plane size should be easily resizable shift+ mouse wheel or similar.

FINAL THOUGHTS

Despite all the inconveniences,  gitches etc, this is a moment I waited for so long for someone to finally care and merge this two technologies properly. Once polished this can be a gamehcanger.

Future
Once all is settled it would be nice if it started to add scan related features such as segmentation. But also I wish there where implementations like semantic segmentation on 2D images so that mascks can easily generated, removing cars, people, sky mask for edge rim fix etc.

If mesh is generated from point cloud it overwrites the mesh generated from depthmaps, may be should be a warning or option to not overwrite rather than having to duplicate beforehand
« Last Edit: November 01, 2022, 09:45:05 PM by Paulov »

RHenriques

  • Full Member
  • ***
  • Posts: 228
    • View Profile
Re: Agisoft Metashape 2.0.0 pre-release
« Reply #37 on: November 02, 2022, 02:06:19 AM »
I follow Paulov in the suggestions and in the idea that this new version can be a game changer in the merging of LIDAR/Photogrammetry.
The features that are also welcome and work very well are the DEM editing features. Very simple implementation and very effective result.
I've already corrected an old survey that had several water pools in the middle using this feature. Very handy.
Best regards



 

Chani

  • Newbie
  • *
  • Posts: 25
    • View Profile
Re: Agisoft Metashape 2.0.0 pre-release
« Reply #38 on: November 02, 2022, 04:30:03 PM »
I did some testing with this and one of the things I found was that when importing a point cloud, it does not replace the existing point cloud even if you have that option checked.
It just seems to add the newly imported cloud as a new point cloud and keeps the old one too.

RHenriques

  • Full Member
  • ***
  • Posts: 228
    • View Profile
Re: Agisoft Metashape 2.0.0 pre-release
« Reply #39 on: November 02, 2022, 05:44:15 PM »
The "Build point cloud" step is very very slow in this version and in 1.8.5 beta. Takes ages to build kernels to all available eGPUs and all the processing is very slow. A Point Cloud that takes 69 seconds to build in 1.8.4, in this version, after 3 minutes, is still in 14% and stalled.

Cheers

PS: In the case of the 1.8.5 (15259) the depth maps stage does not pass 8% and Metashape can only be closed used "Force Quit". 
« Last Edit: November 02, 2022, 07:34:18 PM by RHenriques »

Grischa Hahn

  • Newbie
  • *
  • Posts: 9
    • View Profile
Re: Agisoft Metashape 2.0.0 pre-release
« Reply #40 on: November 02, 2022, 07:28:08 PM »
I am not able to export the Dense Point Cloud in e57 format with Point Attributes "Color" and/or "Normal". Other formats work fine here. Version 1.8.4 works also fine.

RHenriques

  • Full Member
  • ***
  • Posts: 228
    • View Profile
Re: Agisoft Metashape 2.0.0 pre-release
« Reply #41 on: November 03, 2022, 04:03:08 AM »
The problem I've talked about before may be a problem from the MacOS version only.
Version 1.8.4 is working fine here but the most recent ones do not.
I'm using external AMD 6800 and 6900 XT. It can be something related with that too.
« Last Edit: November 03, 2022, 07:03:24 AM by RHenriques »

Paulov

  • Newbie
  • *
  • Posts: 44
    • View Profile
Re: Agisoft Metashape 2.0.0 pre-release
« Reply #42 on: November 04, 2022, 02:16:59 PM »
Color point cloud export is glitched for .las and .laz.

I was able to export colored dense point cloud as .pcd

____________________________________

Hability to colorize lidar point cloud directly is AMAZING!! I thought it was  mandatory to create a mesh!
____________________________________

Despite on IMPORT and EXPORT I input no shift, it seems like coordinates of the LIDAR data are messed up on export since I can´t alignt it back with the original.
« Last Edit: November 04, 2022, 02:30:45 PM by Paulov »

Paulov

  • Newbie
  • *
  • Posts: 44
    • View Profile
Re: Agisoft Metashape 2.0.0 pre-release
« Reply #43 on: November 06, 2022, 12:22:22 AM »
I discovered there is classification !

The ground clsassification works fine, it takes forever tho.

One glitch I found is that if the cloud is already classified with ground and you classify again with new parameters it seems not to work, until you reset classification.

Also I find a bit confusing when you've dense point cloud and scan cloud is not clear on what cloud classification is going to perform.

The semantic classification did not work properly at all on my scan. I'll try again in other clouds.

----------------------------------------------------------------

When you mesh a point cloud, the vertex colors are very poor, with the densities the meshes have and the quality of the coloring initially the cloud has it could be much better.

Bzuco

  • Full Member
  • ***
  • Posts: 226
    • View Profile
Re: Agisoft Metashape 2.0.0 pre-release
« Reply #44 on: November 06, 2022, 11:05:08 AM »
I wanted to know if you have been able to solve the problem that processors with many cores have, in the tests carried out by Pubget Systems https://www.pugetsystems.com/labs/articles/Metashape-Pro-1-7-2-CPU-Performance-Roundup-2261/  it is clear that a ryzen 7 5800x (8 cores) has more performance than a ryzen 9 5950x (16 cores) which it doesn't make any kind of sense, from what I saw the problem was seen in the tile model but there have also been problems with the dense point cloud, it is assumed that the 5950x will tend to take less than twice as long as the 5800x or something approximate but no is the case, even the ryzen threadripper 3975wx(32 cores) is slower than the 5800x. I have done my own tests but only on my laptop that barely has 4 cores, so I could not draw any conclusions about it, it would be very helpful if you can answer that question, and the last question would be, how recommendable is it to use memories DDR5 in metashape, are they worth it?

5800X/5950X/3975wx  has default TDP limit 105W/105W/280W.
TDP limit is causing lower clocks on higher core count procesors. E.g. 105W on 5800x means, that during processing that procesor can run on ~4.0-4.4GHz, but the same limit on 5950x means, that clock are lowered during processing, way below 4.0GHz...from that reason build dense cloud task does not scale linearly with core count. It is even worse for 3975wx, because 280W limit is not enough for high frequencies for all 32 cores(just ~3.0GHz or lower).

The solution is unlock cpu power limit in motherboard bios and set properly PBO settings and also CPU core voltage. Then all higher count processors will scale up much much better...but I am almost 100% sure that this kind of fine tuning settings was not case in puget testing methodology....everything was set on default.

Lower CPU core frequencies are also causing lower frequency of memory controller which increases processing time.
DDR5 can speed up a bit processing, but if they are more expensive than DDR4, buying better CPU or GPU make more sense.