Using the railway network as a flatbed scanner

People mostly loved the project’s spirit. The write-up landed because it is a classic "use X as Y" hack, playful enough to be art but technical enough to teach something real about image formation. The useful context was that this is not a weird one-off technique. It is the same basic idea behind photo-finish cameras, many industrial conveyor inspection systems, pushbroom satellite imagers, and railway track inspection rigs. Once motion is part of the camera, image quality depends on how well you know that motion. The strongest technical comments sharpened that point. A stationary line-scan camera looking at a passing train is relatively tractable because the subject speed is close to one number you can estimate. The inverse setup in the post is nastier because the camera is moving through a 3D scene. Nearby objects sweep past faster than distant ones, turns change the geometry, and no single correction can remove fringing and distortion across all depths. That is why several people suggested adding more sensors or reference imagery, like an IMU, raw GNSS, or even a second conventional camera, if the goal shifts from stylized output to accurate reconstruction. Others surfaced practical tooling around the niche, including custom viewers for absurdly wide images and standalone line-scan rigs built around embedded GPUs. A smaller but recurring point was that the concept is already mature in production settings because those environments control speed, viewpoint, and lighting. That is also why some readers were unimpressed by the aesthetics alone. The train-window version is valuable less as a route to perfect pictures than as a vivid demo of what breaks when the world is not constrained.