Revo Scan - Weird texture artifacts & far too long texturing process

Hi there,

Since the last update of Revo Scan (v4.0.3.0705c) there are occurring some weird texture artifacts on the scanned model and besides that the texturing process takes much longer than with the previous version.

For this test I’ve used the following scan settings:

And here you’ll find the video of the scan test:

As you can see in the video, the texturing process has a total time of about 9 minutes. The CPU load increases up to 100 percent within the first half minute of the process and keeps staying there for about 3 minutes before it settles between 20 to 40 percent for the next 5 minutes. The CPU then gets a little peak load of 83 percent at 8:22 minutes and then decreases again until the process is finished. After that the scan has overall weird texture artifacts and ugly/incorrect color gradients.

Everything was working fine before I updated Revo Scan.

Maybe someone’s here who can help me? I would appreciate it!

Please let me know if you need more info.

Thanks in advance!

1 Like

Thanks David , I changed the CPU multi threading and disabled it and now I am getting the same results as you do , it may be related to the CPU and the system .

There are delays in the last software because the firmware and algorithms got upgraded to deliver better scanning results , and I find the pitch point Is more dense than it should be at 0.1mm with POP2 what can result in longer processing time than usual .

But the Texturing process is killing it , not nesesery and too long in your case . So I am wondering if the CPU threads got stuck while processing . Are your using multithreading or not ? You can check it in your bios .
When I set multithreading on Auto things processing much faster for me and the CPU never reach even 100%, usually at max 67% and idle between at 12-17% , but after disabling it, things getting definitely slower … for some it helps for others it don’t .

Texturing will be fixed in the next build so we can forget about it at this moment as the problem is already addressed.

1 Like

I haven’t changed anything on my computer since updating Revo Scan but I think that multi threading is activated as usual but I’ll take a look in the BIOS as soon as I can. Maybe an update for the BIOS will fix the issue, which I haven’t done since a while.

If the multithreading is ON in bios , disable it , and if disable turn it ON…
Then check again with quick scan if anything improved, on some system the multithreading set ON improve like on my system on others it works better when is set off .

1 Like

Usually it don’t in this case if it was working just fine with your system since the begining , unless the updates of the Bio indicates any changes in the way your CPU is working .
If changing the multithreading do not helps anyway , there may be something else on the hand .
I saw so many people having issues then after fresh installation of Windows everything was working fine again with their system .

Also it would be great if you stop overlapping your scans while scanning , and scan simply 360 degree rotation only , fuse it, mesh it and texture and record it .

Sometimes multiple overlapping delays the whole process as 80% of the whole scan is removed in the process , this also delays the Texturing as not all the color frames are usable …
So making simple 369 rotation will show the true process of your system and the software .

We are all lazy sometimes and we don’t want to merge objects from multiple scans , but that is actually the way of scanning professionally.
Revo Scan can handle the whole scans in one process and clean it up correctly , but it can take some time , and each system handle it differently.

1 Like

So I checked the BIOS and SMT is set to Auto. I won’t change anything there as it was working before without any issues. For now I’ve updated the chipset/CPU driver as this maybe can improve some things, There’s also a BIOS update available but I’ll try a scan first before flashing the BIOS and then I’ll try a scan afterwards but only if the texturing process will still take that long on the first run.

Okay, I’ll do just a 360 scan but Revo Scan should be capable to notice unnecessary color data and remove this data before the texturing process to speed up the process. Why should it keep the color data if it isn’t usable anymore?

Well, I think that more people wouldn’t have problems with that way if the workflow would be more comprehensible but for now I find it very complicated switching between Revo Scan and Studio with all the back and forth just to get a nice and clean 3D model.

Isn’t this what Revopoint promoted their scanners and software to be able to?

I won’t do a scan anymore now because it’s 2 AM here and I’m tired but I’ll promise to upload the video in the next 12 to 20 hours so please be patient with me :wink:

1 Like

If you are ok with only 25% of the quality then yes … scanning the whole objects at once will make you lose some of the sharper details one way or another if the object you scanning have the details to begin with .

I am definitely not doing it for my work scans , but for quick showcase and tests definitely, it is up to you what you do with that .

Looking forward to see the difference after your updated the driver , what can be the key here , as that is important update .

And of course the 360 degree scan in color , I want to see how your system handle that so I can give you some more ideas .

1 Like

Just a tip, you don’t have to mess with bios to change multi-threading. Go to Task Manager then the DETAILS tab and you can right click on anything in there and set the AFFINITY to either all, just 1 thread, 2, 8, etc for the max of your cpu. Mine only goes to 16 threads (and ryzen 7 1700) so 0-15 shows up and I can just select which I want to use.

1 Like

Multithreading is about programming, you will still multithreading with just one core.

Setting multithreading OFF will still run all cores , and each core will have only one thread in place of multiple .

For example 8 cores have 16 threads , with no hyperthreading you will run 8 cores and 8 threads in place of 16 .

Setting multithreading off will avoid the program delays and waiting for other threads to process the data if this is the case , but not all systems and programs profit from that .
That why having it on Auto allows some programs to take advantage of it if needed.
My system profit from multithreading when using Revo Scan , as the CPU never get to 100% while processing , the multi threads do the work faster .
On other systems it don’t .

My CPU isn’t that fast anymore, I guess I’ll have to do an upgrade ^^

Here’s the scan I made after updating the Chipset/CPU driver but this time I only did a single 360 scan, so comparing it with the other scans doesn’t make sense. Anyways, it processed a lot faster :sweat_smile:

For you it maybe make no sense but for me it do , it seems like it works exactly the way it should regarding your CPU power.
When you capture so mant frames , you capture also so many pictures that need to be aligned to each frame that was used while fusing, later the pictures need to be aligned with each face of your mesh so this process take some time in most programs .
Professional programs transfer RGB colors data direct to textures/UV but Revo Scan use pictures to generate UVs what is much longer and separate process from color per vertex data , and how fast your CPU, RAM and M2 SSD is, how faster the process .

You also need to use proper pitch point setting regarding your scanned object . The bust you are scanning don’t have higher resolution than 0.2mm , so scanning it at 0.1mm will only waste your resource and time . Sometimes less is more .

I believe the driver update did made some changes in the process , but avoid over scanning multiple frames as it is just waste of your time and resource and your scans my be less sharper than scanning just one round at a time , fuse it and continue scanning .

But Revo Scan seems to recognize if a face of the scan is overlapping in most cases and cleans up the model if possible. In this case Revo Scan should also be able to delete or clean up the color data.

Which workflow is better?

This sounds like professional experience but I guess that not every user will have this knowledge at the beginning. How do I find out what pitch point I’ve to set for an object and why is it possible to set it to 0.1mm if this isn’t recommended fot the POP2?

I cannot confirm this yet but I’ve ordered a new CPU already and then I’ll do some more testing when it’s installed.

I don’t know what you think about the current workflow within the Revo software but in my opinion it’s not very user friendly and intuitively. I’m working on a suggestion to improve the workflow an the UI of the software for a better overall user experience.

That is not what I was talking about , I was talking about the time that your computer spend cleaning all the overlap frames you create while scanning . You are not taking a video of the object, you capturing the objects faces, different angles of the same faces will create overlapped frames that need to be cleaned and it take time . So you can make your life easier if you do it the way you should and not following your common sense, as common sense is not common after all :wink:

It is not about what is better , it is about what are your skills in 3D modeling and what you can do, the automatic texturing is for people that have no clue how to do it or want to spend time learning other programs . In professional work the textures are not usable for anything the way they are generated at this moment , but improving are in progress so it may be better especially for MINI .

Not really professional , but for anybody that is interested in 3D modeling or at least want to learn the basic to better understand what they are doing . There is no scanner in the world that make a perfect scans for you by a click of a button , this services are very expensive as it is time consuming work , POP2 or MINI provides so good results it almost cut my work in 3-4 weeks .
You can follow the Auto settings that Revo Scan offer, if you have no idea what settings you should use , setting everything on high will not always deliver what you want . So Auto settings will be the answer, and I tested it and works very well especially for fusing , saving lots of time .

Looking forward to it , remember that the speed of your RAM also make a huge difference here and of course as always M2 SSD what in my case it write down with a speed over 2250 MB/s a data the size of 1GB compared to regular SSD at over 500MB/s and Normal Hard Drive at 140MB/s , it also depends of what size it handles, how bigger the size how slower it will be.
So everything is connected here .

I don’t have issues with that , everyone has own workflow , there is no one workflows for everyone , as everyone think differently, and have different kind of system to work with , as they say no all dogs names are REX . You need to devenlop your own workflow , as I mention before the Auto settings are very user friendly and everyone without knowledge can follows it , 3D Scanning is like 3D Modeling , you need to evaluate your object, make plan how to approach it, if you are looking for a click button and done , you will not find it anywhere no matter how much money you spent on scanner, how more expensive how more difficult it is to produce proper scan , as scanning is just a half of the medallion, you need also be able to process it to some level .

I had many suggestions regarding the UI since last year, but nothing was really taken in consideration , not even dark mode , beside the buttons names :laughing: nothing was changed so good luck! Suggestions are always welcome … as long it don’t downgrade the scanning quality again as it happening so many times already . You see one software build has perfect and fast point cloud, the other has wonderful textures , but somehow nothing come together into one build yet … it is always something missing or get broken on the way . So be careful and know well what you wishing for … :wink:

BTW I tested human torso scan on my new Android octa core 8GB RAM tablet and the fusing was faster than on your first video with the bust , I did overlapped some areas on purpose , but you see Android app has superior point cloud fusing with POP2 that I am missing in Revo Scan since February …

1 Like

To answer you more clear , look the image below , left is MINI at 0.02mm and Right is POP2 at 0.1mm
do you see any differences ? not and the files size is also almost the same , becouse there are no details resolution below 0.1mm to be captured better with MINI so the results will be the same , maybe more smoother with MINI becouse if Blue Laser technology .

You can really measure the details on your scanned object and evaluate if you need to go to 0.1mm , are there details on the object smaller than 0.1mm ? or maybe only bigger than 0.5mm … this way you can evaluate what setting you need to use , the Auto setting evaluate the value pretty good .

It is like you have a model with a 0.02mm details and you print it at 0.3mm … everything that is below 0.3mm will be not visible in the print … and same goes for an object that is 0.3mm in details printed at 0.1mm , you can’t get it better looking as it is already is .

No meshing setting will make your model looking better than the original fused point cloud you created, and each edit will only lower the quality and set it apart from the original , that are some rules that you need to know .

Same goes for people that scanning an object that have not finer details than 0.1mm saying MINI and POP2 are scanning the same way as they see no improvements, well it can’t produce more details than the original object have right ? Scanning a chair at 0.6mm pitch point will not give you more details than scanning it at 0.1mm , but the process will be 5 times faster.

So next time you scan something , measure the details and set the proper pitch point , to save your time and computer power.
I use digital caliper for measurements if I am not sure .

2 Likes

When I first got my Mini (Beta tester), I tried scanning my girlfriend’s Hummel-type figurines. That was a mistake as they are very smooth.

Hummel figurines have very low details in general, propably around 0.3mm , good for POP1 :wink:
The busts are not better …

But shouldn’t this been done right at the beginning or after the process of fusing the could points? If so, then this shouldn’t influence the texturing process.

This is the way Revopoint has promoted in most of their videos :wink:

Well, I would say that I know a lot but definitely not everything about 3D modeling but I think the color data from the scanner (besides the current issues) is quite useful, especially for photogrammetry specific content. The biggest flaw in this case are the UVs as they generate way too many unnecessary islands.

My RAM shouldn’t be a bottleneck for now. I’ve 32 GB DDR4 @3200 MHz and I think this should be sufficient for the time being. I can only run one M2 SSD on my motherboard and this is reserved only for system related processes and software. Sure, it definitely would speed up the scanning process if I would use more powerful hardware components but for now I’ve to deal with my current system.

No, I was talking about the workflow a user needs to do if he not only wants to clean up the scan inside Revo Studio but also keep the color data. We already had a conversation about it and you told me how to achieve that but this didn’t work at this time because of a bug. Do you remember? However, this is what I meant with insufficient workflow.

In my opinion they probably have to do a complete rework where every software not only comes together in one single cross-plattform application but also gets a fresh and potential restart as for now there are too many quality differences between each Revo software for each platform e.g. a user who only has a smartphone isn’t able to calibrate his Revopoint scanner with Revo Calibrate or to tweak the model with Revo Studio. Besides that, Revo Scan then should only be used for scanning and Revo Studio should be used for everything else upon that. Just my idea but I guess this would improve the workflow and makes it more comprehensible for users.

Thanks for this amazing tip and for the detailed explanation. Now I’ve learned and understood more about the pitch point. I was already thinking about buying a digital caliper because of 3D modeling.

That is not how it works , you have 2 types , RGB color per vertex that capture RGB data while capturing frames at the beginning , and then you have UV’s texturing produces what is a separate process .

That is how you do things if you want the textures to be optimized , optimized textures use every fragment of the UV’s so no, it is not unnecessary , it could have even more islands for higher quality .

It should be ok, however remember you don’t use 32GB of RAM with Revo Scan but what is left after the system used , and creating overlapped scans can get heavy on the RAM , with the new firmware and software the RAM usage is slightly higher but it should be enough , if not you would crash the software propably .
So what you use for Revo Scan if not M2 SSD ? what the point to running only the system on it, if the system can’t perform the task at optimal speed ? you could just use it and move the files once a week to your second HD to safe space .

It is not a color data , you mean textures and UVs … color data is safe when edited with Revo Studio , UVs and Textures are not . I asked already the Development team to add Texturing function to Revo Studio but the problem is that after the model is changed , or reduced it is impossible to retexturing it using the old data , the model structure changed … the only way is to create textures from the color vertex data what is in my opinion much better looking after with less artifacts and issues .

That is complicated , you can’t have the same quality on all platforms as the mobile devices are not powerful enough to do the task , so phones or tablets are limited , however you can still scan pretty good point cloud and mesh it in Revo Studio for top quality . Other way it is a toy on a phone and can’t be better as it will eliminate many devices automatic as it use RAM for processing , you buy 8GB RAM device , the system use 2.5GB and you have left 5.5 GB for scanning , I would not even expect Revo Studio to be available on mobile devices unless you want to wait 24h for each process …

If we don’t have what we like, we have to like what we have lol.

The development team is very busy right now and until they adress all bugs reported , there will be nothing new added. The software is very sensitive and do the job as the firmware allows , any small adjusting can easily broke things as it is happening at this moment , so all we have left is little patience and let them do their job so we finally have a solid foundation for new stuff .
Any changes to the software requires changes in firmware so it is not so easy task when you deal with all the customers devices that need to be upgraded…