Video Sculpture + Nothing: The Art Of Illusions
Panoptes Mirror
Panoptes Mirror is a circular video infinity mirror which live streams your eyes in the mirror when you step in front of it. The mirror is a circular infinity mirror containing a round screen with a resolution of 448x32 pixels. A webcam with a light is attached to the bottom of the mirror. The webcam feed is analyzed by OpenCV using dlib to identify your eyes which are sent through syphon into Touchdesigner and then streamed to a raspberry pi controlling the led matrices. When there are no viewers in front of the mirror, a peacock feather themed animation plays on the screens enticing the viewer to approach. When the system detects that the viewer is present, peacock feathers which are instanced inside of the mirror transition into the viewers eyes.
CONCEPT
Panoptes Mirror is a commentary on surveillance capitalism and the implicit consent we give to have our faces and eyes scanned (usually unwittingly) as we navigate public and retail spaces. The piece is inspired by the Greek legend of Argos Panoptes: a 100 eyed giant tasked with guarding IO, A consort of Zeus, by his jealous wife Hera. Zeus had Hermes slay Argos who was rewarded for his service by Hera who embedded his eyes into the feathers of the peacock. Peacock feathers in the case of Panoptes mirror are a metaphor for the flashy content advertisers use to gain impressions and harvest our attention and data. The feathers are displayed on “infinite screens” visible in the infinity mirror, a parallel to the endless feed, when someone is attracted to the mirror, the webcam light turns on and they become aware that they have now been captured by the mirror and their eyes have now been embedded into the peacock feathers just like Argos.
Keeping the work running:
This is hard because it take a lot of computing power to run openCV + Dlib + Syphon + Touchdesigner + Processing running for an extended period. If this was full time installation, I think I would have to purchase a computer that is built for intense processing for a longer duration or figure out how to move all of the work onto the raspberry pi. I think it would be possible to record the peacock animation and use clever masking in opencv to instance the feathers and transition them into eyes but that is a larger project for the future. I am also unsure how long the raspberry pi can run led screens without overheating.
Picasso’s Dreams
I have been looking for a way to bring general adversarial network art into the physical world since starting ITP and this project finally inspired me with a way to do so. I chose the Cubist movement of art as my inspiration and created a 2.5D sculpture out of CNC’ed plywood based on a “Two Women On A Beach” (to the right).
The reason I chose Cubism was because there are direct parallels between the Cubist painters trying to reduce complex figures and scenes into minimalistic features that when presented next to each other allow the viewer to see something from every angle at the same time and GANs which learn the features that make up a set of images representing an idea and display that idea in innumerable ways.
I painted the sculpture white on the faces and black along the edges of the ply which creates a somewhat illusory effect as you walk by it in real life.
For the content, I created a dataset of 1000 Cubist “portraits” - or what I deem to be portraits by cropping bodies and faces out of cubist works, then upscaled them if necessary to 1024x1024 pixels.
I then trained StyleGan2-ADA on 1 GPU over the course of a couple weeks to produce models that can both generate more abstract textures as well as discernible faces.
I am very lucky to live with a creative tech who works in the event business and seems to have most of the equipment available in the ER. My mapping setup involved three projectors at different angles to hide as many of the shadows created by the depth of the sculpture as possible. If I were to remake this project for an installation, I would figure out my projector setup beforehand then reverse engineer a sculpture that was not an incredible pain to map. I used Izzymap in Isadora and a tripplehead2go hdmi breakout in order to control all the projectors from one computer. My projector setup was not ideal for preventing drift when cars drove by with a foam cooler on top of the carpet.
Along with the content itself, the placement of the layers of projection on the sculpture where inspired by cubism. Generally, each layer on the Z-plane contains content from one “latent walk” / video mapped across different surfaces at the same depth with an exception for the head and top neck being similar subjects seen from different angles. If you watch the video closely, you can see that the top plane (arms, chest, & hands) match as well as the bottom plane (back most side of leg and bottom neck), and the next most deep plane (middle leg and bottom of foot).
Untitled Proposal For An Installation
Collaboration with Nikhil Kumar
A portrait of the “Self” - a proposal for a larger scale implementation. All video content used in this project was shot by us. The system uses openCV + dlib face tracking to trigger the viewers face to be mapped onto a mask on the wall when they step into a spotlight in front of the webcam.
Dichroic Infinity Light Sculpture
For the first assignment I created a light sculpture based off of a generative patch that I created in touchdesigner which was exported as an svg then first plotted to assess the correct dimensions on a cameo vinyl cutter.
Next I laser engraved the pattern on the back of acrylic mirror and applied dichroic film to the front of the mirror. When the plate is backlight, an infinity mirror effect is created along with changing colors both in the z axis and across the x and y axis when the plate is tilted. I tested .125 and .25 inches and decided the thicker the better as it increases the infinity effect.
Next I CNC’ed a light box which contains led strips behind an inset piece of translucent white acrylic for diffusion and finally the dichroic infinity mirror plate and finished it with shellac before painting. I was not able to finish the glue-up / mounting because I had to quarantine due to covid exposure and left my apartment.
