Shopping by Inspiration

What if you would be able to take a photo from an item that interests you most (that cool bag you saw on the street or in the television)?We developed a quick prototype capable extract meaning out of the images by using AI-solution (e.g. take a picture of a cake and AI-will tell you it’s cake and do a search against your product catalog). Below you can find some initial insights. For example, the image we used to search below can be found here.

Capturing image search through mobile is quite an easy and fast way to process product search, especially when EAN- or QR-codes are not available. Besides, it can be done in any location without bulky bar-code scanners.

Delegating the product search for AI interpretation may enable better understanding of the product’s itself. For example, AI may be used to recognize not only what type of fruit the customer just shoot (Apple), but also what cultivated variety is in question (Ambrosia).

More interesting solution areas would be to understand contexts of images in more detail. For example, how the AI would work when different types of spaces are being photographed (is the photographed room stylish, what types of furniture it contains, what type of interior is in place), or, when the AI is used to interpret something (how many apples is there on the basket, and should an order be placed automatically when there’s too few of them).

Although this was very interesting rehearsal, more work would be needed to deepen the AI’s capability to understand the image context’s. The first thing you need to do is to sample and train the AI to understand the context of the items you specified. The accuracy of training depends how accurately you want the AI to recognize the items (in our rehearsal the cake was recognized as a 92% accuracy). The more you train and wider dataset you have, the more accurate results you’ll get.

Sensos

Comments are closed.