# Datalab Cup4: Unlearnable Datasets Imagenet — Data

来源：https://www.kaggle.com/competitions/datalab-cup4-unlearnable-datasets-imagenet/data

本地 Meta Kaggle 快照；不是本次实时页面。

# Dataset Description

In this competition, we're classifying 2 types of animals based on their images.

# Files
This competition provides its files in Numpy array format. The samples and the corresponding labels are stored in two separate files. For instance, the label of `x_train_imagenet_unlearn[0]` is `y_train_imagenet[0]`.

- **x_train_imagenet_unlearn.npy**: training samples.
- **y_train_imagenet.npy**: training labels.
- **x_val_imagenet.npy**: pre-split training samples.
- **y_val_imagenet.npy**: pre-split training labels.
- **x_test_imagenet.npy**: test samples. You will be predicting what classes of objects these fall into.
- **sample_submission_imagenet.csv**: a sample submission file in the correct format.
    - `id`: a unique ID for each test sample. The shape of `x_test_imagenet.npy` is `(2220, 224, 224, 3)`. `x_test_imagenet[0]` represents `id = 0`.
    - `label`: predictions of images represented by the sample.