FoodLog: using computer vision and social networking to support dietary assessment

Recording daily meals is useful for personal healthcare and medical treatment procedures. However, almost all existing dietary assessment methods are challenging for various reasons. People find it hard to make a detailed record during meal time, but might not remember the details later. Further, expert assistance is needed to derive nutritional information from the records. We propose an Internet based system called FoodLog for easier dietary assessment. To record a given meal, a FoodLog user photographs it using a digital camera or camera phone and emails it to the system. Image analysis determines the nutritional composition and the calorie content of the meal and logs the results. The user can view this log in different formats and visualizations (e.g. calendars, graphs, maps), and edit the data to correct any mistakes that occurred during image analysis. They can also browse logs of other users to find similar information about different types of meals. FoodLog was launched in April 2010, and had 1934 registered users during the first six months. They uploaded 31318 photos of meals. Work is in progress to add functionality such as recognition of specific meals and automatic meal recommendation.