This is not an apple

This is not an apple


For Your Eyes Only

(previously called Dark Objects)

Or, when is an apple no longer an apple?

This project will explore the aesthetics of visual obscurity in an era of machine learning and computer vision. The result of this exploration will be the production of a series of objects for human eyes only: objects that cannot be recognized by computer vision, but can still be recognized by humans.

For example, in the image above I am clearly holding an apple. However when I submit this image to a Google Image search, the results are obscured. It no longer appears as an apple to algorithms. By contrast, an image of a conventional apple yields the correct result.

Test: Is This An Apple?

To illustrate this concept I used a conventional apple and covered one side with dots, leaving the other side unaltered.

The Google search results for each object are shown below.

Project Goals

In this project I will use machine learning algorithms to build a computer vision profile for recognizing an object. Since the profile constitutes a definition of what the object is, the profile can also be used to define what it is not.

The first For Your Eyes Only (FYEO) object trial will be borne from a conventional apple. However, instead of the conventional interpretation of the apple as a forbidden fruit, this FYEO object will represent biological evolution towards anonymity and visual obscurity, camouflage from machines. A possible trait that I imagine could be useful in the future, when cameras and machine learning are even more ubiquitous.

In this way, visual obscurity can also be a type of camouflage or deception, which historian Roy Behrens notes “has always been critical to daily survival—for human and non-human creatures alike—and, judging by its current ubiquity, there is no end in immediate sight”.

There is additional support for the philosophical value of visual obscurity in Baudrillard’s statement that, “[t]he moment a thing is named, the moment representation and concepts take hold of it, is the moment when it begins to lose its energy”.

The goal of this project is to create something in between, an object that can be both detectable and obscure, evolving new visual mutations that satisfy human perception and override machine vision.

Production

To build the FYEO object series, I will be creating machine learning algorithms to generate visual mutations of objects that are robust to object recognition. The final objects will be 3D modeled and printed.

Presentation

The final 3D printed objects will be presented in a custom case accompanied by a collection of images showing the evolution of the objects’ mutations.

Timeline

With support from Rhizome, my proposed project will be executed in a year or less. The following is an estimated timeline for research, prototyping and production:

  • Stage 1: Create object-detection profile for apple using eblearn
  • Stage 2: Develop machine-learning algorithms to deconstruct the object-profile
  • Stage 3: Generate prototypical 3D models to test
  • Stage 4: Fabricate 3D models

COSTS

  • Research: $1,000 (est. 3 months)
  • Development: $1,500 (est. 3 months)
  • Assistants and Developers: $500
  • Production: $500 (est. 1 month)
TOTAL: $3,500

SIMILAR WORK

  • CV Dazzle
    Makeup and hair styling used as camouflage from face detection
    This project uses a custom developed computer vision library built on top of OpenCV to create camouflage from face detection
    http://cvdazzle.com

UDATE May 16: As part of the preliminary research for this project, I’ve prepared a basic reverse-engineered face using a similar method.

Haarcascade Frontal Face Profile: Reverse Engineered Possible Face Image

Haarcascade Frontal Face Profile: Reverse Engineered Possible Face Image

Haarcascade Frontal Face Profile: Reverse Engineered Face Image

Haarcascade Frontal Face Profile: Reverse Engineered Face Image

The FYEO Objects project will adopt a similar approach of reverse engineering an object detection profile, though will extend this concept to 3D forms.

ADAM HARVEY

EDUCATION

  • 2008-2010 Masters of Professional Studies, Interactive Telecommunications Program, NYU
  • 1999-2004 Bachelor of Arts, Integrative Arts, Pennsylvania State University

EXHIBITIONS

PRESS

PRESENTATIONS/INTERVIEWS

AWARDS/GRANTS