PRODUCTION SCENARIO
A vision app must report where each of three damaged packages sits in a loading-dock photo, not just that damage is present.
Which computer vision task fits?
Answering here is anonymous. Nothing is saved unless you sign in.
Show answer and explanation
Answer: Object detection with bounding box coordinates
Object detection is similar to tagging, but it returns bounding box coordinates in pixels for each object found, which is what 'where each package sits' requires. Classification labels the photo as a whole and reports no positions. Tagging lists what is present, including contextual terms that cannot be localized, and still gives no coordinates. OCR reads printed text on the labels rather than locating the packages.