kaggle's been p useful for building a baseline but i usually can't help thinking that unless i spend a significant amount of time getting my own data that i'll just end up w something everyone else already has.
i've been on a bit of a blackjack kick and will probably build a detector with yolo at some point. i'll likely have to scrape for card imgs and then take my own. there's probably a bunch of pretrained models online but it'll be a fun weekend side project.
saw this thing on twitter where ads at stadiums are tailored to customer depending on location, lots of theories on how they do it not sure if any are accurate but its p cool https://twitter.com/haltakov/status/1415009284958703620