Hacker News Reader

Comment by umangsh | Hacker News Reader

OpenFoodFacts offers variety of options to consume their data: https://world.openfoodfacts.org/data. From my experience, their coverage for packaged foods from various countries is amazing, though it can be fairly noisy.

In US, USDA provides a more curated dataset through FoodData Central (https://fdc.nal.usda.gov/) - smaller dataset but higher quality (nutrient coverage, number of samples tested, etc).

View on HN

I played with both, and was indeed surprised with how good the USDA dataset was. For most scenarios, you can get away with even their "Foundation Foods" database. Users can use the foundation foods to approximate various meals, and it usually has better accuracy

Yes, USDA foundation foods is very good for estimating calories/nutrients by ingredients, perfect for getting caloric content and nutrients of a recipe.

Nutritionix is very good for packaged foods, they got both an API and an app which can accept barcode scans.

I’ve also played around with plain google searches “calories <food> <x> g” works great for many simple items if all you care about is calories.

Foundation foods data includes few hundred foods and absolutely sufficient for meal tracking if you cook at home and measure portions. Another cool thing about the data set is that USDA publishes individual samples from across the country used to derive foundation food nutrition data, for e.g https://fdc.nal.usda.gov/fdc-app.html#/food-details/1999632/..., which helps visualize changing nutritional trends for foods with location and time (https://news.ycombinator.com/item?id=31229335).

I'm a big fan of NCCDB.

http://www.ncc.umn.edu/ndsr-database-page/

Interesting, I haven't come across this dataset before. What do you like about it? How does it compare to USDA or OpenFoodFacts if you've used either of them?

Reply on news.ycombinator.com