This reading challenges the common stereotype that computer vision has to function similarly to human vision by highlighting how opaque digital images are, meaning that computers cannot interpret them without algorithms. It shows that computers basically cannot function if there were no algorithms to program them. Unlike humans that can constantly interpret context, different types of objects and different meanings, computers rely on simplified techniques, algorithms and systems such as frame differencing, background subtraction and brightness thresholding to detect movement or presence. For example, the reading explains that frame differencing compares pixel changes between frames, while background subtraction depends on differences from a stored image of the scene. This actually made me realise that computer vision is actually much more dependent on certain conditions having to work than I really initially thought. For instance, background subtraction only works if lighting and contrast are carefully designed. This shows that seeing for computers is more about structured assumptions. Meanwhile, the reading shows that we can actually guide what computers see by designing the background and physical environment, such as using high contrast, stable lighting or even infrared to improve detection. This connects to the idea of interactive media, where it’s not just coding systems but also designing physical and visual conditions that shape interaction.
What I felt was most interesting was how computer vision’s ability to track people overlaps with surveillance, and how artists actually engage with this. The reading discusses works such as Sorting Daemon and Suicide Box, where tracking systems are used to monitor and analyze people, sometimes raising ethical concerns about profiling and data collection. This made me question whether interactive art using computer vision is actually empowering participants or not. On one hand, projects like Videoplace create an embodied interaction where the human body becomes part of the system, but on the other hand, surveillance-based works turn viewers into objects of analysis. I think this is what makes computer vision powerful in art as it can both engage and critique. However, the author seems somewhat biased toward presenting computer vision as accessible and empowering for artists, while not fully addressing issues such as privacy and ethical risks. This raises a question: as computer vision becomes easier to use, how should artists balance creativity with responsibility? I also wonder whether using computer vision in interactive art always requires some level of surveillance, or if there are ways to design systems that avoid this dynamic altogether.