Hot take: watching the self-checkout cameras at my Kroger in Boise made me get why AI vision is actually hard
I stood there for 10 minutes last Tuesday watching the little screen flag my bananas as a 'possible theft' three separate times while the machine also somehow missed the loaf of bread sitting right in the bagging area. It hit me that these things are doing real-time object tracking on 20 different items at once with weird lighting and people's hands flying everywhere, and we still yell at them like they're dumb. Anyone else work with computer vision stuff and know why weight sensors plus cameras still can't tell a zucchini from a cucumber?