You are working to create an object detection system, like the ones described in the lectures, to locate cats in a room. To have more data with which to train, you search on the internet and find a large number of cat photos.
imho, The answer could be, “We can add the images to the train set.” Even if this is a localization problem, we can train the system first to detect cats (classifications) and use that model in the first layers to refine it for localization further.
That does sound like a plausible answer, but the quiz questions are sometimes a bit subtle and you have to read and carefully consider all the answers. The other fundamental point to consider is whether the question is a “one answer only” or “check all that apply” question. You can tell by the shape of the “click targets” on the answers: if they are round, it is “one only”. If they are square, it is “check all that apply”.
We aren’t allowed to share quiz questions and answers on a public thread, but you can take a screen shot and DM it to me.
To close the loop on the public thread, we had a private conversation about the particular question and the problem was the additional subtlety that was missed was that the question was asking about object detection and localization instead of simple image classification. It’s not just that we need to recognize that there’s a cat in the picture, we need to draw a bounding box around any cats that are in the image. Of course that was one of the major subjects of C4 W3, which is where the quiz question occurs.