third prototype - angle definition

Aiming to leave the face area without any physical barriers I decided to try another prototype.

Second protyping - angle definition

The goal of this prototype making was to find out which angles work with the mannequin head.

Fast prototyping e-makeup device

In order to control the angles of the projection on the eyes, I decided to create a metal skeleton of the device.

Next step: Attach the mirror and figure out the angle, and then move on to the 3D reconstruction.

Drafts e-makeup device

Inspired by parasites and the relationship between hyperfemininity and hypermasculinity I doodled some ideas for the device.

Photogrammetry final

I tried to use my old computer for trying out the kinect sensor because I found out Microsoft hasn't developed a driver for the Windows 10. My old computer is too slow to work properly so I wasn't able to get any good results. So I kept doing the photogrammetry, this time on Agisoft Photoscan.

First I tried again with the pictures my boyfriend took from my face.

Regular reconstruction

Summarized reconstruction

As you can see, I couldn't get any good results, so I tried to do new pictures with the mannequin head.

Agisoft photoscan reconstruction 1

Agisoft photoscan reconstruction 2

Agisoft photoscan mesh development

Cinema 4D Mesh

Maya - Mesh cleaning

Maya - Mesh simplification

What I learned: After trying many times to recreate my face on 3D I find hard to believe that is easy to recreate something that moves without the proper equipment. From now on I will use the mannequin head as a base for my design.

Photogrammetry 4, 5 & 6th try

For this new step, I took 82 pictures rotating the camera around the subject (my face) instead of rotating my face around.

After that, I decided to use the pictures without any crops and medium reconstruction quality on Colmap

The result was better than before, especially when the cameras were created for the reconstruction, but still unsatisfied I tried the same process on Zephyr 3DF

Dots on Zephyr 3DF

Mesh on Zephyr 3DF

Mesh .obj on Maya

As you can see the result using Zephyr 3DF was still bad, so I decided to crop the images highlighting just my face.

Wrong move, actually the reconstruction turn out worst because the software couldn't read the background.

So I decided to go back to Colmap and try the highest quality of reconstruction with the raw pictures,

High reconstruction process 1

High reconstruction process 2

Reconstruction on Meshlab

 Clean reconstruction on Meshlab

Mesh extraction

Mesh cleaning

Mesh smoothing 1

Mesh smoothing 2

Mesh .obj on Maya

Mesh .obj on Maya 2

Mesh .obj on Maya 3

What I learned:

1. Pictures are reconstructed better if they have a constant background point.

2. Colmap highest reconstruction understood the back of the head. 

Photogrammetry 3rd try

Watched a good tutorial online and the guy was using Colmap + Meshlab, so I decided to try them

out:

Colmap 50 photos Reconstruction

Colmap reconstruction map

Meshlab dots reconstruction

Meshlab mesh reconstruction

Cleaning the dots

Mesh reconstruction after cleaning the dots

What I realized: Watching most of the tutorials users usually take pictures not rotating the object, but actually rotating the camera around the object. I will try that next time.