[rev087]

The bit about taking a picture of food you are about to consume and reading/logging nutritional data might be closer to viability once smartphones get Kinnect-like 3D sensors, making it possible to read volume data. Google's [Project Tango](https://www.google.com/atap/projecttango/) is one such initiative.

Does anyone know of any computer vision research into identifying food?

[jxf]

This will be an extremely difficult thing to do with just computer vision alone. Small changes in food preparation and recipes that don't yield very visible differences in the final product can have huge impacts on nutrition.

For example, a slice of chocolate cake prepared with non-caloric sweetener will look much the same as a cake prepared with brown sugar, but it'll have significantly fewer calories. (It'll also taste gross.)

[cma]

Sounds like a job for Mechanical Turk