Memento (Beta) : 3D Model Creation from Photos or Scans

You probably have seen how real life objects (or people) are scanned and turned into 3d models using high tech equipment, and maybe more than once you thought your life would be a lot easier if you could just scan that clay model instead of modeling it in your application of choice. There have been many attempts to provide users with the technology to turn photographs into 3d models, with different results. Now, Autodesk is offering their technology to do this with the release of Memento (currently in beta).

The workflow in Memento is very straightforward: you take many photos of your object, upload them to your cloud (part of your Autodesk membership), wait for them to be processed and then download your 3d model. Uploading the photos to the Autodesk server means you don’t do the conversion on your computer, but rather the software on the servers does it for you. This is a nice thing because it means you can continue working instead of wait a long time for the model to be extracted. The workflow is simple, but it is not easy.

The first thing you need to know is that taking photos is very tricky. First, you need to make sure you lock the focus, so nothing in the picture is blurry. Then, you have to make sure your object takes around 70% of the picture space, so you need to take pictures at a somewhat close distance (using zoom is not an option). You also need to use a tripod so you make sure your camera doesn’t move, and lastly you need a lot of diffuse lighting with almost zero shadows (no direct lights, no three-point lighting and such).

 

I found that it is good to go around the object taking the pictures, at every 30 or 45 degrees, at different heights. When you move from one height to the next, make sure you take “in-betweens” so it looks like the camera is “transitioning” from the higher position to the lower position. The entire idea is that you need reference parts of your model when transitioning from one picture to the next, so the software knows how a specific picture relates to the previous and next images. Basically, this is like simulating the camera taking pictures as it moves around the object over and over in a smooth path. This also means you can’t just take pictures at random, one from the front, then from the back, then from the side, then from the front-above, then from the side-below…

You can then take specific closeups to get the details, also following the same rule I described above (no “jumps” from one side of the object to the other). When doing this, keeping the focus locked is important, because certain cameras try to refocus as you get closer (or completely lose focus). You can take up to 250 photos, and that’s usually more than enough.

 

Editing the Mesh

Memento will usually reconstruct not just your model, but the entire scene you shot, so your final model will show all the back walls, props, objects and such. You can then use the editing tools to remove unwanted geometry (basically you select the faces you don’t need and delete them), so you’re only left with your model. Models generated by Memento can be very high poly, so you can use the software to reduce the polygon count to a number you’re more comfortable with.

 

The software also has tools to set the object size. You can place a ruler next to the object you’re shooting, and then use Memento to resize your model to match its real-life counterpart. This can be very useful if you need to have exact dimensions in case you plan to use it in arch-viz, or something similar. Memento also generates textures based on the photographs, so the resulting model is a textured duplicate of your original (textures will also replicate the lighting conditions available when you shot the object).

Memento can be a real time saver if you have a physical version of the object you need in 3d, because remeshing a high polygon object is a lot easier than building it from scratch.

As I said before, the resulting model is extremely high resolution, and if you want to use the model for animation or games, that’s not very good. You can reduce the polygon count inside Memento, but I think the best option is always to use Maya or TopoGun to retopologize your model, so you can control how your faces and edges flow along the surface of the model and avoid any deformation issues along the way. Since the new model would have a different set of UVs, you’d also need to transfer your textures from the Memento model to your new low-poly model, and while you can’t do that inside Memento, you can do it in Mudbox, for example.

 

One Complaint

If I had one complain about Memento it would be the lack of documentation. They have some videos and webinars on their website, but I think those are not a replacement of a well-written user manual, specially when you’re looking for a specific answer or way to do something.

Memento can be a real time saver if you have a physical version of the object you need in 3d, because remeshing a high polygon object is a lot easier than building it from scratch. You should definitely give Memento a try so you see what the software can do for you.

 

Notes

Autodesk's Memento is still in Beta (as of 4/11/15) and is free to download. Be sure to read the system requirements as Memento is a Windows-only application at this point. Memento is part of Autodesk’s Reality Computing portfolio and is great companion to ReCap. For more info, click here.The Autodesk video below shows the basic workflow for scanning and creating a 3D model using Memento (beta).

 

 

Sergio Aris Rosa [nemirc], is Sr. Staff Writer for the Renderosity Front Page News. Sergio discusses on computer graphics software, animation techniques, and technology. He also hosts interviews with professionals in the animation and cinematography fields. You can follow him on Twitter, and if you want to see what he's up to you can visit his blog.

  

April 11, 2015

Please note: If you find the color of the text hard to read, please click on "Printer-friendly" and black text will appear on a white background.