This project compares artificial intelligence (AI) generated content labels for a set of images against identically generated labels for reduced resolution versions of those same images. This project has two purposes. First, in the spirit of Data TRIKE, I am demonstrating how modifications to data can have unexpected impacts on analysis. Second, this is a pilot project to explore data created using the Google Cloud Vision content label feature, in anticipation of a potential wider scale project.
It’s also just fun (and disturbing) seeing the world distorted through the eyes of AI. Feel free to skip ahead and check out the Analysis section first for examples of AI labeled images – some sensical, some strange.