Teaching Machines to See (Part 1): Why Vision Is Hard

Human brains effortlessly recognize objects through attention, memory, and context, but computers see images only as numerical matrices of pixel values. This 2D representation of a 3D world creates fundamental ambiguities, particularly depth ambiguity, since a camera captures light position (x, y) and intensity but loses distance (z-axis) information. The article explains how cameras form images through lenses focusing light onto pixel sensors, and why interpreting these raw numbers remains a core challenge in computer vision.