問題文
A manufacturer photographs each finished part and wants a model that says which of seven known defect categories a photograph shows, or that the part is clean. The photographs are stored as files, not as rows of numeric columns. Which description of this work is accurate?
選択肢
- It is unsupervised segmentation, because the model has to separate the regions of each photograph before it can name a category, and the names are attached afterwards.
- It is regression, because the model has to output a defect severity, and severity is a continuous quantity.
- It is supervised multi-class classification on unstructured data, because the label is one of several known categories and the input is an image rather than tabular columns.
- It is binary classification repeated seven times, and that is the only correct formulation because a single model can output only one probability.