KMNIST-49

The 'KMNIST49' dataset is a benchmark used to evaluate unsupervised federated learning methods, containing images of handwritten digits from the Kuzushiji-MNIST dataset, specifically focusing on the challenges of decentralized data distributions.

Papers using KMNIST-49 (2)

Federated Clustering: An Unsupervised Cluster-Wise Training for Decentralized Data Distributions2024CatFedAvg: Optimising Communication-efficiency and Classification Accuracy in Federated Learning2020