Textured Mesh has low resolution

Why has the Textured Mesh a mutsh lower resolution.

Here are to Images to may explain it better.

Textured verion of the Mesh

Just the mesh without texture. It has a lot more details.

By using Vertex coloring the mesh resolution remains as the original.
The difference between Textured mesh 185Mb to origial mesh 2.5Gb.



Because textured mesh don’t need that high resolution , the textures you are producing contains all the light in it so the details don’t really matters as all you use with it is just ambient light shader . And it is easy to share on Internet .

Plus the textured model view use smooth normals preview .

On top , did you ever created a UVs for a 2.5 GB model ? An island UVs ? ( nobody do that ) … it would take so long on Android system maybe one hour or longer … for that reason the model is simplified to speed up the process .

What you should do is use the low polygon textured model with the UVs , together with the high resolution mesh , and project the details to it … then you can extract the normal maps and displacement maps and have a rendering or game ready prop with the specific maps included .

It is a great workflow , saving lots of time


Hey :slight_smile:
I am realy sorry that I get somtimes complicated and unwanted. Even my Wife hates me sometimes for this. :shushing_face: But this is another Story.

Because textured mesh don’t need that high resolution , the textures you are producing contains all the light in it so the details don’t really matters as all you use with it is just ambient light shader . And it is easy to share on Internet .

Okay sorry, I think I disagree with you on this. High Texture details are always welcome to generate aditional Normal Maps, AO, Roughness and Albedo. Why I can’t decide myself what mesh resolution I want to have? There still is a decimation option inside Revoscan. Funny part is that with Vertex Coloring I get an Unwrapped Model in nearly no time but just with Vertex Color’s. The calculation time for the Texture is nearly the same. Projecting Texturs and coresponding UV’s are mutsh faster in calculation time as generating propper layouted UV’s. Thats why all Scanners and Photogrammetry Software have this chaotic and unsorted UV Layout. I am not so mutsh involved into Internet things. There is no need to finish the Scan inside of the Miraco.

On top , did you ever created a UVs for a 2.5 GB model ? An island UVs ? ( nobody do that ) … it would take so long on Android system maybe one hour or longer … for that reason the model is simplified to speed up the process .

Well yes :slight_smile:. Some Scans I get from the scanning department have up to 25GB, and I have some Workstations to handle this task. With new Domain registrations and PDG’s I splitt the Project across the Network.

What you should do is use the low polygon textured model with the UVs , together with the high resolution mesh , and project the details to it … then you can extract the normal maps and displacement maps and have a rendering or game ready prop with the specific maps included.

Okay I think you knew that this does not work with a Crumped pece of Bred, and other things. You can’t reproject a Iso-Surface like mesh with a LowPoly UV.

The main Problem is that this device may have a 64MP Camera implemented, but in Highres scanning mode it is cropped to 3MP. In all scanning modes I never have seen a 64MP Image, max was 12MP. It is stored as a 8Bit JFIF/JPEG BGR-Format with bad compression, does not have Pixel Binning ore anything else to get a better Texture Quality. There are no Polarization Filters to get rid most of the Highlights/ Reflections. The Isocell HP2 Sensor costs for OEM’s about 50-70 Bucks, and delivers 200MP with a very good SDK to handle most of the problems. For shure the Processor inside of the Miraco cant’t handle 18FPS at 200MP; not even 8FPS. But, after scan is done, you may can do the Pictures, and get use of IR-Depth + RGB to fit to the scanned Object.

Curretly the RGB Cam does not make any sense. 64MP Cam for a Low Poly Mesh, with a blochy Texture is for me realy disappointed.

Please dont get me wrong! The scanning quality is just great, and enough for the team. The problem I have is to send another Team to make Photogrammetry just for the Texture.

Still good, but I am missing the Hirez Texturs! Skin is blurry at all :frowning:

Enough for now…
Maybe I used to stand too close to the wall when I was swinging as a child. :rofl: :joy:

N8, Gerrit


Ohhh, i read over this…
I sometimes are so negativ…
I should write also about positive things…
I do Tomorrow, promise…
And there are a lot of good things… :slight_smile:

1 Like

Gerrit , you are not negative , but I discarded my reply to you because that would not ends well :joy:

But I have some answer for you

First of all the sensor is 48MP as I tested the system myself … as with all Android systems a 48MO sensor need to be pixel binned to 12MP as that improve the light condition that the sensor needs to avoid noises , this sensor is still very much light hungry .

Now going to 3MP , do you have idea how big your project would be of you produce 15 frames pets second at 8K for 2 min ? Almost 300GB , more than the system have space , it is impossible to take so many pictures as it would only slow down all the process and the actual scanning would be affected .

The Single shot at this moment offers 4K photo data and also 8K textures , but when you create 8K textures fir example of Human , the textures don’t need even 4K of photos as it will be rescaled down to so small fragments that it is not even needed anyway . You will need at least 32K textures then you can use 4K photos on that with huge improvement.

MIRACO is not a DSLR camera shooting movies in a 8K with 15FPS consistent of 4 min … and scanning at the same time .

Don’t forget beside color data there are also depth data so it is already 45 pictures per second at this moment …

Be realistic … it is not a photogrammetry, and you don’t know exactly how things works here behind the scene …
Would be that possible otherwise with a powerful PC ? Yes if the scanner would relate on the PC GPU, CPU and RAM but that would make the scanner dependent on a powerful PC and not standalone anymore …

It does have pixel binning because the system indicate it do …

So Gerrit , the resources are limited on Android system , and I will not continue this conversation regarding this subject , I can’t change it , I am not responsible for that , there is not better solution yet …

But still you can create even better results with proper lighting and settings , remember that when you scanning a small objects in color , they do not even take 25% of the main frame so much resource is wasted here , and since the priority is still 3D scanning , we can’t make it working worse just to have some pretty picture …

No professional will need this kind of picture or textures anyway … it is worthless in today’s pipeline the way it is .

I am working for over 25 years with texturing in 3D … I was across the globe with my creations , my students working in biggest studios today , I seriously can’t waste my time on arguing because someone need to put a regular picture on a 3D model and fake half of the work that other professionals have to do to archive the best results on today’s standard .

It is like a flash back to the 90"s here … people need to move on .

Use the great resources they have in place to faking details with a picture , converting it to Normals and AO , like for real ? It is 2023 … nobody professionally do that …it is outdated old workflow .

Sorry Gerrit , but I can’t waste my time on that .

1 Like

Ok, Thank you.

That’s all you offer me back ? :joy:

Gerrit , I suggest you use Single shot , it will give you the maximum quality and really nice results that not one Revo Scan scanner can do in the past .
Of course I beg Dev.team to allow us to use the data image format of 8K on single shot in place of 4K , since it is very helpful with scanning a small objects because for scanning big objects it is really enough and efficient with 8K textures, but for now it is the max possible , however it would be great to have it at least on scanning via PC mode ad tyr computer can handle it … it would increase tue size to 300Kb per image what is not that bad for a single shot since there is enough time between shots to load the data .

I do have the same sensor as Miraco on one of my Phones at full 8K , but it takes around 2-3 sec to process it after taking a shot so for continuous shots it can’t be used … but for single shot yes .

Hopefully when the MIRACO system get even more stable , we would be able to have that as I did not gived up yet on that .

I have found my ways how to make perfect textures with Miraco and a single shot feature including cross polarization that removes all reflections from the light on the surface , but about that more in my next tutorial .


Hi Cathy,
my short answer is, because I don’t want to waste your time.
Sorry to say, and with all respect, but you are not the only white rabbit here.
I am now 32 years in this Business, and still are.

I have seen the Live tonight, and sorry for placing the same questions there. Most of the questions are cleared, or at least RevoP. seem to realize most of my concerns with the DEV team.

As it seems, RGB Cam is not pixel binned for now, and there is a lot to do.
I am really sorry for getting this forum involved in my harsh criticism about the problems I have. I think I will go directly to the support center without a public audience.

This Live was very helpful. Thank you Cathy for your time :slight_smile:


1 Like

Gerrit can you please summarize what in the Q/A was cleared. For beginners like me?

Oh I don’t mean this way , my mining was about having some specific 25 years of texturing and 3D modeling experiences, I am graduate of visual arts colors, light design and a body form. Trusted Google photographer and HDR VR specialist .

It is pixel binning since while testing I had a cess to the whole system and run my android system diagnostic , so I know exactly what sensors are used and how .
It is pixel binned to 12MP , this sensors do not operate well at 48MP without proper light , they are light hungry sensors , that why they need pixel binning to get rid of noises … of course the Android system do not have any additional camera software to bring out the fine running and dynamics that sensor needs .

I used the sensor via the system directly , I have the same sensor also in one of my phones at full 48MP …

Trust me Gerrit , I am a Beta tester of a prototypes , I shredded everything down to the last sensor and gived all my suggestions back to the team including the sensor and what it needs .

You may just send your own suggestions and feedback , but I know exactly what is happening and how it works this way and not another .

It can be even better since it is not yet full potential here … so hopefully improving coming with time as usual .

One thing what will not change is the resolution of data images in continuing scanning mode , as increasing the size will affect the speed of frames and automatic tracking in a results. Already tested out .

The single shot resolution is 4K for data image what is already more than enough for a 8K textures , remember that you do not use the full 4K image while capturing a very small object , it works great with bigger objects , sadly small objects use maybe 1-2K from the sensor , since there is not optical zoom in that type of sensor .

In photography you need to change the lenses , on a phone you have 4 different types of cameras for different situation , you can’t scam macro with a landscape sensor , the shortest distance is very limited too .

I guess different design would fix it like fir example using 2 RGB cameras each for different distance mode to solve this issue for once …

So you see Gerrit … I run through everything for a long time as that is also my priority.

For now some tip …very important …use good lighting at least CRI 95+ , lower will create dirty effect .
The colors will be bright and beautiful.
And of course best option is single shot for excellent color results and resolution, so far triggering single shot via PC gives me the best clean shots …

I may have better advantage here because I use it almost for 3 months and learned well what to do and what to not do .

It is the same with scanning , people pushing the sliders to max for 'better" accuracy and resolution …but don’t know the principle of modeling killing their details in progress …

And don’t get me wrong I love to talking to you , always interesting that why good conversation is about …

And yes according to the Android system the 48MP sensor is pixel binned to 12MP , from what the MIRACO software use 4MP in a single shot function .

Using 8K on a single shot will delay the processing dramatically … even on a very good PC , being there done that .

P.S you never waste my time Gerrit !

1 Like

hey hi gerrit, i was also dissapointed by the texture i got from the miraco and also the fact that image texturing will delete a lots of detail of the mesh. I figured out at least some workarounds that will provide some better results, first of all, you want a model with as much vertices as possible, so fuse at the maximum level not the suggested, skip all other steps for the pointclound like point merging (equalize the density, dont know the english term) and mesh at maximum setting, create vertex based textures and export the model, go back and delete all changes on the mesh and reconstruct at a resonable polycount, thats the mesh you will bake the textures onto, in most cases the mesh is fine in structure when use vertex colouring as well, but when the mesh has less vertex the texture will have less detail as only mesh vertices and not all original captured points are used to generate the mesh (sth that should be possible to change from revo, use pointcloudcolouring instead vertex colouring). also i did a modification, mentioned above, i added crosspolarisation onto my led fill light and also onto the rgb lens (90° angeled) to get rid of most of the reflections. than its important to set exposure and white balance to manual so the colours will stay consistent during the scan. I got good results on medium sized objects by this techniques (10-50cm in size) smaller objects are not as good, but for this putv gave an explaination i can understand as we have to stay quite far away of the object. singleshot mode always offered the better results, especially when scanning inside a lightbox. i anyway prefer to use photogrammetry for the texture. i requested some features that will make this workflow more easy like export the images taken along with the scan including exif data so they either can be edited in photosoftware or be used to align a additional dslr set correctly onto the scandata without having to do it manually, also i requested tag detection, like those circular ones on the calibration boards which are detected by most photogrammetry software as well, and have the ability to export the positions of the detected tags in revoscan so they can be used as groundcontrolpoints. also a detection of tags would improve tracking and recommence a scan building better tiepoints than a pattern of markers.

Hi @bau-3d.ch this method is only good if you create textures from color vertex , the points are important as each point represents a pixel in your textures later , so you want your mesh to have a as much as possible points , however generating mesh with higher count of point than the pointcloud will not add any additional pixels , it will just clone the points …

The textures created from color per vertex can be as good as the pointcloud , so fusing the pointcloud at the highest fusing pitch point is the way , however it will not generate proper clean mesh …or can be meshed at the same grid settings meaning you gonna lose more than 1/3 of the point after meshing anyway …

So whatever trick you try , it will not works at the end .

What you should do is :

Fuse the scan at highest settings
Clean the lose points and overlapped points .

Export the point cloud and mesh it in Cloud Compare at level 12

In my time fuse the same scan again at 0.2 pitch point in Advanced mode
And mesh it after at the proper Grid 0.20 under Meshing tab .

Now export the meshed scan from Cloud Compare and replace it with the one you already meshed in Revo Scan …
Restart Revo Scan

Load the project again and now you have mesh that was produced in CC ready to be textured using color per vertex at maximum pixels ratio .

This is only the way to trick Revo Scan to get higher results with textures by color per vertex .

If you want to use full quality of Miraco textures , use single shot … it will give you 4K per image ( frame) and 8K textures .

The texturing of the model using image data works best if the model have lower resolution . As with all 3D programs in the world , you create UVs on a reduced polygon model , to save the processing time and resource .