Saturday, 24 January 2015

Design Iterations - Face Tracking Project

Face Tracking Project - 8 

Following on from placing my work with the space (see blog post Face Tracking - 7) I have since found some ways to rectify some of the issues that I was encountering. One of the first problems that I had when conducting the test in the space is that the camera was struggling to pick faces up. At first I thought that this may be the quality of the camera which I was using. However I have found that it was in fact that camera location and angle which was the problem. As seen in the image below I have place the camera higher up, so that is more at head level. This means that it can more easily detect a persons face. This has improve the performance of the installation in many ways. The first being it can pick up the users much easier which means that unlike when it was begin tested in Weymouth House the user could keep a natural position instead of having to get close to the camera. Another improvement that this new camera position has had on the installation is that the installation is less "glitchy". What is meant by this is that where the camera can now better detect a persons face the system can track the face better. What this means is that the circle stays on the screen better and does not randomly cut out. 

Higher camera position improves performance of the installation 

Another issue that I had encountered was the fact that the black background did make the TV look like it was off, when there was no users interacting with the installation. This was a simple fix as I have made the background a grey colour or from 0 to 50. This means that the installation does not make the TV look turned off, as well as still being dark enough to make the colours pop. 
Changed background

The last issue that I had with the project, I have not been able to rectify. This problem was when there are no users interacting with the work there is nothing to entice someone to go and use it. I believe that having the circles move around the screen by themselves and then when a face is detected then track them as it already does would be a good fix for this.

No comments:

Post a Comment