Make Menus and Recipes More Discoverable with Image Search
Food is such a wonderfully diverse world. In the sense of taste, and in the sense of look. When it comes to digital space, where taste is not expressed effectively, discovering food through dish names is not enough. There should be a way to explore this world through how they looks. Let's see how image search makes discovering menus and recipes much easier and interesting.
Scenario 1 - When you don't know the name of a dish
You'll be amazed by number of dishes in ordering apps or cooking apps. BigOven says they have over 500,000 recipes in their database. Those who can remember recipe names and categories are admirable for their memory and rich knowledge of cuisine. But not everyone can do so. Many of us see a type of food somewhere, but cannot tell its name.
Foodie apps and websites should make it easier for their users to search for dishes with images. Users want to find what they see, not find the name of what they see.
Scenario 2 - Finding alternatives that look and feel similar
When we look into a food image, what we like is the feeling that the image arouse in our imagination. When we look in an image of egg tart, we may crave for the taste of crispy and mildly burnt skin. Or we may crave for soft, sticky and sweet egg cream. Unfortunately, we are not in a situation when we can order or make egg tarts. Can we tell the search engine to find us something similar? What we often do is to look up in our mind similar dishes, and type it in search box. Now, visual search can help us with that, because foods taste similar are likely to look similar.
How image search for food works?
Foodie websites and applications have tons of images. Therefore, it is not difficult to enable image search. I make a demo of with Lookuq from random food images, training with our standard model. Check it out!
To understand how I do it, watch our tutorial video showing the steps to create data for automatic machine learning.
If you think your websites or mobile apps should have image search, we'd love to help. Questions and ideas are welcomed too, drop me an email (huong.nguyenthimai [at] sharitek [dot] com).