Skip to content

Dataset · apple

Apple labeled images for federated learning

apple/flair

A large corpus of labeled images for benchmarking federated learning (FL) models.

What's inside

The dataset contains between 100k and 1M labeled images with text tags. Created by the Apple team to evaluate federated learning algorithms, it was presented at NeurIPS 2022. The data reflects real-world distributed environment characteristics, and ready-to-use benchmarks for TensorFlow Federated are provided.

What you can do with it

  • Benchmarking federated learning (FL) algorithms
  • Training and evaluating image classification models in a distributed environment
Tasks
Image classification
Size
100K–1M rows
Downloads this month
785
Likes
23

How to load it

from datasets import load_dataset

ds = load_dataset("apple/flair")