Point Cloud Plugin

[phoboz]

@rotacak

@Jay_Dizzla, @: Increasing the intensity helped, but it still looks far from the “Displaz” screenshot:

The scan is in .las format and does have only grayscale colors (that is ok). You can check the scan file there. Density is ok, that can be adjusted, just that missing contrast (?) is the problem.

First, sorry for the delay.

It appears the file is encoded with intensity in 0 - 4095 range instead of the 0 - 65535 as per LAS specification, hence everything comes through very dark and lacking contrast. What software did you export it from?

[rotacak]

First, sorry for the delay.

It appears the file is encoded with intensity in 0 - 4095 range instead of the 0 - 65535 as per LAS specification, hence everything comes through very dark and lacking contrast. What software did you export it from?

I got is as is from the scanning company. So I don’t know, but I can try to ask. Will be possible to display it correctly - probably stretch it to 0 - 65535 by plugin? Or by different way?

[rotacak]

@**** I got reply from the company. Point cloud is from this program: TerraScan (http://www.terrasolid.com/home.php). But that range can be different and it is set by postprocessing when data are downloaded from the scanner (0-255,0-4095, 0-65535,…).

[phoboz]

@rotacak

@**** I got reply from the company. Point cloud is from this program: TerraScan (http://www.terrasolid.com/home.php). But that range can be different and it is set by postprocessing when data are downloaded from the scanner (0-255,0-4095, 0-65535,…).

Thanks for the reply.
Currently, the plugin supports 0-255 and 0-65535. I haven’t come across 0-4095 till now, but will add it to the list
The easiest solution for you would be to convert to one of the other two ranges

[Scarfab]

Hi !

I’ve find a solution for the lower density pointcloud when camera is close to point cloud, it was due to manual focus distance, now when i set it to 0 all the cloud is visible without lower density

Now i have a last problem…

How to reduce or disable this side line kind of effect on video, it seem that is motion blur effect but its strange because i have disabled it and set it to 0 everywhere ( projet setting, posprocessVolume, camera etc )

thanks in advance

https://forums.unrealengine.com/core/image/gif;base64
​

[phoboz]

@Scarfab

Hi !

I’ve find a solution for the lower density pointcloud when camera is close to point cloud, it was due to manual focus distance, now when i set it to 0 all the cloud is visible without lower density

Now i have a last problem…

How to reduce or disable this side line kind of effect on video, it seem that is motion blur effect but its strange because i have disabled it and set it to 0 everywhere ( projet setting, posprocessVolume, camera etc )

thanks in advance
​

Hi,
Glad you found a workaround
The effect you are describing could be caused by the TXAA antialiasing - try changing that and see if it helps

[Scarfab]

Hi,
Glad you found a workaround
The effect you are describing could be caused by the TXAA antialiasing - try changing that and see if it helps

Yes its from the Antialisasing… I’ve tested diferrent value with :

r.DefaultFeature.AntiAliasing=2
r.TemporalAACurrentFrameWeight=0.2
r.TemporalAASamples=4
r.Tonemapper.Sharpen=1

but i have always blurred ghost lines…

have you some tips or recommanded setting for the AA ?

Thanks

[phoboz]

@Scarfab

Yes its from the Antialisasing… I’ve tested diferrent value with :

r.DefaultFeature.AntiAliasing=2
r.TemporalAACurrentFrameWeight=0.2
r.TemporalAASamples=4
r.Tonemapper.Sharpen=1

but i have always blurred ghost lines…

have you some tips or recommanded setting for the AA ?

Thanks

Temporal AA will always produce those ‘ghosts’ due to how it works. Switch to a different AA mode or disable it altogether

[david.gillespie]

Haha, mobiles are not on the immediate roadmap

Just bumping the iOS support as it’s a use case very pertinent to us

[anonymous_user_81a537a1]

I have a problem in the 0.8 version. I am kind of new to this but I wanted to open a 100 ish GB pts file and it crashes almost instantly. I have opened 7GB file that worked fine. I have 64GB ram and 8GB vram. Is this a not enough system resources problem or are that large of point clouds are not supported yet? Thanks

[phoboz]

@.gillespie

Just bumping the iOS support as it’s a use case very pertinent to us

PM sent.

@anonymous_user_c6357811

I have a problem in the 0.8 version. I am kind of new to this but I wanted to open a 100 ish GB pts file and it crashes almost instantly. I have opened 7GB file that worked fine. I have 64GB ram and 8GB vram. Is this a not enough system resources problem or are that large of point clouds are not supported yet? Thanks

Hi @anonymous_user_c6357811

Due to 32-bit indexing, 4.25 branch has a limit of ~2.1 billion per asset - we have now updated all internal code to use 64-bit so starting from 4.26 there will be no hard cap. However, a 100GB pts should be under that level anyway - please check the task manager for RAM usage when importing. If you hit the limit, you may need to increase your page file size.

Alternatively, try to split the file into a few smaller sections.

[david.gillespie]

Thanks @ - On managing point cloud data for other applications - I believe a collision mesh can be generated from the plugin from a point cloud. Would there be any way that a textured version could be generated via similar means - that way a point cloud could be downsampled to a textured mesh which would be useful in cases where there are lower powered devices or where the plugin hasn’t been ported? The mesh approach wouldn’t supplant being able to work with point clouds however I image this would be a really useful feature in a lot of use cases as the asset would be a lot smaller than 6,12…36Gb point clouds in a built project.

[david.gillespie]

Any news on e57 support yet? We have one case where we have a unified e57 scan which is made from 29 separate scans. Whilst we have the separate files as LAS/other ASCII formats, loading in all 29 can be time consuming plus it would have excess data (where the single e57 is ~1.6Gb and the non-unified LAS files are 29 X ~600Mb LAS = ~17.4Gb).

I know the import of all those clouds and placement could be automated however that wouldn’t solve the additional excess data from the non-unified data.

(Let me know if I should submit something on UDN if it would help to talk specifics)

[anonymous_user_63759055]

couldn’t you convert the merged e57 data to LAS with a tool like CloudCompare and then import with the PointCloud plugin?

[david.gillespie]

Yes you can with CloudCompare as that’s what we’ve been doing. However when you open the e57, as it stores the original scans, when you export you still end up with each scan as a separate file (well I assume without additional processing in CC). e57 is a standard across the industry so that would be our preferred direction.

[phoboz]

Hi @.gillespie

Sorry for a late reply, I was attempting semi-vacation

Re Meshing
It is on our todo list, but it’s unlikely to make it to 4.26 due to time constraints.

Re E57 support
We still have it in a prototype phase, but I believe it should make an appearance in 4.26 release

Hope that helps

[david.gillespie]

Hi @ - No worries - everyone deserves a break!

e57 is exciting to hear as it makes things more efficient for us. Still interested in whatever iOS support there may be as otherwise whilst it is possible to mesh point clouds, the results can be a bit larger and more cumbersome than one may like for mobile.

[samdax]

any option to make it work in Android?

[phoboz]

@samdax

any option to make it work in Android?

Not currently. Your best bet is to use Pixel Streaming.

[david.gillespie]

Hi - I’ve been seeing a lot on On-Demand streaming of point cloud data with 4.25 but struggling to find the documentation/examples. Any suggestions on the best place to look as runtime loading of point clouds we’d like to test out!