Classification of a photo using machine learning tools can be challenging. Over the last year, significant improvements in the algorithms that power these tools have dramatically improved their efficacy.
Microsoft recently partnered with Vectorform to build an Android application surfacing a food’s nutritional value based on a photo of the food.
The main functionality is powered by Microsoft’s Custom Vision Service, the younger sibling of Microsoft’s Computer Vision API that allows you to fine-tune a predictive model to the dataset at hand.
“With just a few clicks and no custom code, you can easily build your own predictive model using CustomVision.ai, a new addition to Microsoft Cognitive Services. The Custom Vision Service makes image classification powered by Deep Neural Nets very accessible to developers.”
“Custom Vision is suitable for a broad range of domains. For example, you could use this technique to detect items in a customer’s online cart, recognize UI elements, pre-filter images into categories to simplify further analyses, and so on.”