Week 3 Quiz Yolo object detection picture

I just find the quiz is full of grammar test? it is testing grammar not concept…
Why some questions are like english tests?

why this sentence below is correct?
True. Only one of the boxes for a car is kept because the score of 0.73 is higher, and also because it has a larger bounding box.

Semantic segmentation can only be applied to classify pixels of images in a binary way as 1 or 0, according to whether they belong to a certain class or not.
Why this statement is wrong?

Isn’t semantic segmentation generally used in a multi-class context? If so, you’re outputting the predicted class, not a binary flag. To the best of my knowledge, when the problem reduces to the binary case it is explicitly called out. Sometimes it is used to differentiate something (any one of the label classes) from nothing (background)

Are you asking about the grammar of this sentence? Or its technical merit? To me, a native English speaker, the sentence makes an assertion (only one of the boxes is kept) and then provides two justifications for it. First, that the score is higher. Second, that the bounding box is larger. I don’t see any real problems with it from a grammar perspective. Technically, I like it less. Hard to be completely sure since I can’t see the quizzes (no current subscription) but it seems like this is related to Non-max-suppression (NMS). NMS is used to prune likely duplicate predictions at the end of the YOLO pipeline. If there are two predictions likely to be of the same object, it will keep the one with the higher confidence and discard the other. However, the size of the bounding box is irrelevant. NMS doesn’t have access to ground truth, so it can’t know which of the two candidates is more accurate (that can only happen in training, which doesn’t use NMS). NMS uses bounding box shape (not strictly size) to compare two candidates and determine whether they are likely to be the same object. But it uses IOU as the metric, not size. In my opinion, the second clause also because it has a larger bounding box is not true during any phase of YOLO, neither training nor runtime.