steelwARks is an exploration of trackerless, large-scale augmented reality. My goal was to create a system that would superimpose 3D models of the Homestead Steelworks on top of the Waterfront (the development that has replaced this former industrial landscape). Instead of attaching the 3D models to printed AR markers, I used landmarks and company logos on-site as reference points to position several simple models of rolling mills. When the models are seen onscreen, overlaid on the environment, it gives the viewer a sense of the massive scale of these now-demolished buildings. As I was testing my system out at the Waterfront, I got a lot of positive feedback from some yinzers who were enjoying the senior special at a nearby restaurant. They told me it was very meaningful for them to be able to experience this lost landscape that once defined their hometown.

1st Test

This project was built using openFrameworks. I combined the Fern and 3DmodelLoader libraries, using videoGrabber to capture a live video feed from an external webcam. The main coding challenges of this project were getting the libraries to talk to each other and projecting the 3D model properly. Fern doesn’t have an easy built-in way to attach 3D models to images it tracks so I had to hack my own. I also had never worked with openGL before so getting the model to render correctly was tricky.

The computer vision from the Fern library worked very well in indoor testing, but when I used it outside, it had some issues. I had to update the images Fern was using as the day went by and lighting changed. This is a tedious process on an old core2duo machine, sometimes this process took 10-20 minutes. When using a large, 3D object as a marker, it was difficult to get the webcam pointed precisely at it to register the image. In the end, the program was only stable when I used logos as the markers.