Adnose – Anosmia and (ML) smell prediction

Adnose is an interactive sculpture in the shape of a nose created at NYU, Tisch School of the Arts ITP (Interactive Telecommunications Program) which combines image recognition and machine learning using a Raspberry Pi to predict the smell of any object placed underneath it. – Adnan Aga

I was born without a sense of smell. which is both a blessing and a curse. I haven’t ever experienced the smell of New York in the summer, or the aroma of my grandma’s cooking. Being an anosmic has meant I live a life that is deprived of one of our fundamental senses and have to reorient my understanding of smell. My primary way of knowing that my climbing shoes have the foulest odor is not because I can detect the volatile molecules entering my nose but because of my mom and sister. I’ve built up prior context of what certain objects smell like because of social cues, past responses, and general understanding.

For my graduate thesis at the Interactive Telecommunications program, I focused on creating a system that calls attention to way that I and other Anosmics perceive the smell world. Adnose is my attempt at making that system in the form of a large sculptural nose.

I decided to 3D scan my face, and get a high resolution model of my nose. That was then 3D printed into multiple parts, and then assembled to create a 21″ fabricated sculpture. This was then finished and spray painted for a more sculptural look.

A Raspberry Pi acts as the “brain” of the system. A Time of Flight sensor detects when an object is brought underneath it and then triggers a fisheye lens camera to take a picture. The picture is then sent to Google Images to find the best match for what the object is. Then the subject of the picture is fed into GPT-4 to create a description of what that object smells like. It proceeds to print out a receipt from one of the nostrils of the predicted smell while simultaneously reading the description aloud.

Google Images (Image recognition)-> GPT-4 (Smell Prediction) -> ElevenLabs (Text to Speech)

Project Page | Instagram

Adnan Aga is a Dubai-born multimedia artist based in New York City. Aga’s work in the field of creative technology and experiential design includes a variety of dynamic and unique projects from building giant gumball machines to musical soap dispensers to password protected cookie jars. His professional and artistic practices focus on designing and developing interactive experiences that explore the whimsical and downright comical areas of our lives. He merges cultural touch points and technology to create installations, wearables, software programs, and digital products that subvert our expectations of the role of sensory performance and information in daily life. Currently, Aga is developing a series around the physical experience of anosmia, the absence of the ability to smell, in an attempt to create empathy around smell-related disabilities/sensitivities through the reprogramming of our everyday sensory experiences. Adnan has a B.S in Mechanical Engineering from the University of Pittsburgh, and will complete his M.P.S at the Interactive Telecommunications Program at NYU Tisch in May 2023.