Skip to content

Dataset · EduardoPacheco

Food and ingredient segmentation on 7,118 photos

EduardoPacheco/FoodSeg103

Benchmark for pixel-level segmentation of dishes and individual ingredients in food photographs.

What's inside

Contains 7,118 food images with pixel-level ingredient-level annotations across 103 categories. Images are selected from the Recipe1M dataset and manually annotated and verified by annotators. The dataset is split into training (4,983 photos) and validation (2,135 photos) sets.

What you can do with it

  • Training semantic segmentation models for dishes and ingredients
  • Developing automated calorie counting and dietary analysis systems from photos
  • Comparing and evaluating segmentation algorithm quality on complex objects
Tasks
Image segmentation
Size
1K–10K rows
Downloads this month
2K
Likes
20

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("EduardoPacheco/FoodSeg103")