Computer Vision
The articles juxstaposing of computer vision with human vision was very interesting. While some aspects are similar, such as a common use of types of lenses for both recording video and through retinas, the image processing appears to be where differences begin to proliferate.
An interesting point I found the article made was that videos are inherently recorded to store pixel information, but not necessarily scene information. For instance, a night sky is recorded as lots of black pixels–rather than some encoding of a night sky parameter enabled. This fundamental concept means that complex algorithms must be constructed to reconstruct and interpolate the scene information from pixel values. Furthermore, there are still many video encoding formats (e.g., H.264, H.265), so standardization is further lacking in this regard–introducing additional complexity to the process.
One of the techniques I found intriguing is the background subtraction technique, where an initial reference image of the set is first captured. Then, the reference is used to systematically distinguish which objects belong to the scene, and which do not.
The surveillance art, which monitored the Golden Gate Bridge, sparked considerable reflection. I found the author’s point particuarly pointed, when it was revealed that the art had captured a considerably higher number of suicides than what was noticed through traditional systems. However, I can also see how recording these events is also uniquely invasive to the subjects, who are likely unaware that they have become part of an art piece–and did not sign up to be so. This work was only made possible through computer vision.