Datasets:
File size: 2,454 Bytes
6ed7dd1 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 | ---
license: cc
task_categories:
- image-classification
- feature-extraction
- zero-shot-classification
size_categories:
- 1K<n<10K
---
# Food Nutrients: A Macronutrients Dataset
`food_nutrients` is a dataset of visual and nutritional data for ~3k realistic plates of food captured from Google cafeterias using a custom scanning rig.
This dataset provides annotated pictures of food plates along with **calories**, **macronutrients** (fat, carbohydrate, protein) for the total plate and **for every ingredient** as well.
| Column | Definition |
|------------------|----------------------------------------------------------------------------------------------------------------|
| `image` | a 640x640 top-down image from a realistic food plate in the Google cafeteria |
| `id` | identifier of the sample |
| `split` | the split of the sample, only `test` for now |
| `ingredients` | a list of the different ingredients, with the mass (`grams`), `calories`, `fat`, `carb`, `protein` and `name`. |
| `total_mass` | total calories of the plate |
| `total_calories` | the total mass of the food on the plate |
| `total_fat` | the total fat on the plate |
| `total_carb` | total carbohydrates on the plate |
| `total_protein` | total protein on the plate |
## Source
This data was taken from Google Research's dataset [Nutrition5k](https://github.com/google-research-datasets/Nutrition5k). It's cleaned up and filtered, because this dataset had many issues such as missing images or missing calorie information for certain samples. That's why this dataset is a bit smaller than 5k samples. |