
Using 2 image sets is currently only supported with Agisoft photoscan and Capturing Realitry, NOT with autodesk recap 360. The image set without projection is used to generate the texture (the colors), so that you do not see anything of the projection in the finished model. The image set with projection can be used to render the 3D mesh (the actual model). The projection is "flashed on in 300 milliseconds". On the 1st image set no projection is present, the 2nd image set the projection is displayed.
CAPTURING REALITY MULTIPLE TEXTURE SKIN
This allows you to scan things that have little or no texture, like skin or black leather.Įach camera will take 2 full frame photos in 0.2 seconds. What this means is that you can use a projection system to introduce extra "noise" on a subject. The Pi 3D scanner support scanning with Projection. The total hardware cost (self build), including lighting is around $10,000 / € 10,000. The Raspberry PI 3D scanner is a multi-camera solution for a fraction of the price. As you can imagine, this is a very expensive solution. This has been used in the last few years typically based on 60 to 120 digital SLR cameras (canon/nikon). Using a multiple camera system is not new. If you want to scan a high volume or want to scan things that can not stand still (childeren / animals), you need a solution that can scan the object in a fraction of a second. The downside with most other scanning solutions is that it takes some time to make a 3D scan, from 6 seconds to a few minutes. A Microsoft Kinect camera or an Arctech EVA is an example of this method of scanning. This camera projects many infrared dots onto the subject and uses a camera to figure out where these dots are.

A popular method is by using a "hand-scanner" that typically uses RGBD depth sensing camera. There are many ways to make a 3D scan of something.
CAPTURING REALITY MULTIPLE TEXTURE DOWNLOAD
You can find sample scans in the download section, these include the original images and the 3D Model (.OBJ and texture). These 3D models then can be used to make a 3D print (3D mini figurine), measure body or place the model in a game. With the a set of 100 photos, you can use photogrammetry software/service like Autodesk Recap 360 or Agisoft Photoscan or Capturing Reality to turn the images into a 3D Model, with full texture (full color). Each Raspberry PI unit (Camera Unit) stores the photo for later processing.

It is designed to make a full body scan of a person in a single "shot". For more information, I encourage you to visit the AR Quick Look Gallery page for examples and check out the 'Object Capture' session. Reality Composer makes it easy to combine multiple assets into a single scene and add interactivity to models. The Raspberry Pi 3D Scanner is based on the popular single board computer Raspberry PI and the 5 or 8 Megapixel PI Camera. Object Capture creates high-quality assets that are ready to be viewed in AR Quick Look.
