3D Scanning & Printing

For this weeks assignment we had to scan and digitize an object as well as produce something with one of the 3D printers. For this I wanted to experiment with some of the non hardware specific techniques such as photofly and so mostly focused on scanning.

Z Corp Scanner

Using the handheld scanner developed by Z Corp I was able to digitize the objects with varying amounts of success. As the device uses a laser it struggled to calibrate on the metal surface of the robot and so there are many artifacts produced within the output.

The texture of the knitted donkey similarly meant that much of surface details were missed and so it was unable to produce a clear surface from the data.

3D Printing

This week I decided to finally make peace with Blender and so set out to draw a human skull. After spending a few hours battling with the software I decided that this was wasn't going to be possible with my skills and so turned to thingiverse for an alternative. Having found the potential model I was able to use what basic skills I had developed with Blender to scale it and prepare it for printing.

PhotoFly

One tool being developed by Autodesk is Photofly which is designed to allow you to generate point cloud data from a collections of static images. I attempted to digitize a couple of objects using this software but was unsuccessful in generating any usable content from it.

The first thing I tried to scan was my head which I did by taking around 40 photos at various angles. These were then uploaded to the PhotoFly service where a basic mesh is then returned fro you to inspect and adjust. Sadly though the software was unable to stitch most of the scene together and so required a lot of manual identification of the various elements of the scene.

As it takes awhile to produce a collection of images from which the point cloud data can be calculated I decided to write a quick python script that extracts still images from the video. Using this I was able to submit large quantities of images to the service with the hope of improving the final rendering. Sadly despite submitting many hundreds of images the service was still unable to stitch the content and so all the produced meshes are incredibly course.