Multi Label Dataset Families

I’m working with a data set that has a hierarchy of labels. Product Manufacture; Product Family; Product. So a product only belongs to one manufacture, however a manufacture can have many different products and product families. I’d like the model to predict all 3 labels for each image. Should I approach this as a multi label problem such as the planet example or develop 3 different separate models - one for manufacture, family, and product. Any suggestions?