As we learned on Monday, deep learning-based visual recognition algorithms are not always reliable. Each algorithm carries with it an error rate, regardless of the hype around the capabilities of modern AI. In this homework, you'll have an opportunity to probe the reliability of one such algorithm: Azure AI's Image Captioning System. Find three example images that result in an incorrect or unexpected caption when provided to this algorithm as input. Explain why you think that these images caused trouble for the algorithm. (Hint: there is no need to brute force this process. Think about the various confounds to computer vision that we've talked about this semester.)
Add your three photos (along with the resulting bad captions) to a Google doc, which should also contain your explanations for why they fooled the algorithm.