In the article “Computer vision for artists and designers” Gloan Levin proposes an important idea – the field of human-computer interaction has expanded, including now not only researchers and engineers, but also people from creative industries. As such, the scope of projects and inventions in this area grows as well, especially considering the increasing accessibility of software and hardware that allows experimentation for interactive media artists and even undergraduates.
Computer vision is essential for the machines to interpret the visuals, just as human vision is crucial for us to see what is going on around. However, there is a fundamental difference between these two – humans analyse the picture subconsciously, while computers depend on algorithms. Both are dependent on assumptions, but while human vision can interpret any new unexpected object by relying on their previous knowledge, a computer will likely break down. In order to help the computer to see what we want them to see, certain techniques have been developed to facilitate the machine’s analysis of the visuals – frame differencing, background subtraction, brightness thresholding, object tracking. All these methods allow to set up a path which the computer will follow when it analyses what it “sees”, creating specific conditions for the machine to make its output more reliable.
In addition to that, Levin touches upon the idea of surveillance in interactive art. Thinking about rapid technological development in the last decade, I am always anxious about the emerging tools for surveillance, especially facial recognition systems. Although I understand the necessity behind these innovations, the issue of privacy remains. As I was reading the article, I remembered a video that explained how the AI human recognition algorithm can be fooled – a special pattern for clothing was developed, which confuses the AI when it analyses the image, making it lose the human face. I think this is an interesting area for investigation in interactive media arts – how can patterns disturb or help the computer vision? And what are the boundaries of intervening in the surveillance process?
Skip to content
