Totmaat/howest-animals

Animals (cats / dogs / panda), 64x64

The exact train/test split produced by the datasplit step of the Azure ML pipeline (seed 42, 20% test per class): 800 train + 200 test images per class. Images were resized to 64x64 by the data_prep step.

Layout is the imagefolder format: data/<split>/<label>/<file>.jpg, so the label comes from the folder name. The configs block above maps each folder to a split.

from datasets import load_dataset
ds = load_dataset("Totmaat/howest-animals")  # DatasetDict with "train" and "test"