Choosing a machine learning technique

A fashion retailer maintains a database that contains informational tags for each item of clothing: colour, material, type of garment, product code, price, and a descriptive name. There are also photographs from various angles that are associated with each item. The retailer operates both online and from physical stores, and would like customers to be able to take pictures of items in$-$store or from print adverts and find the corresponding item online.

In your post, answer the following questions:

Which supervised learning technique would you recommend for an image$-$recognition problem? Justify your answer by mentioning the characteristics of this technique.

If data about sales and warehouse stock was also available, what other business problems could you solve? Which techniques would you use to solve these problems?