# UW DATA 558 Spring 2019 Competition 1 — Data

来源：https://www.kaggle.com/competitions/data558spring2019-competition1/data

本地 Meta Kaggle 快照；不是本次实时页面。

# Dataset Description

<p>The training set consists of 500 images from each of 10 categories of animals, from lorikeet to goose (see <a href="https://gist.github.com/jovianlin/5fb426c38fddfdf9a699a8824450c7dc">here</a>, but note that the integer labels are not the same as ours). The validation and test sets each consist of 100 images from each of the 10 categories. The images in the training and validation sets are saved in folders by category. </p>
<p>We are providing you a Python 3 script called "extract_features.py" that will extract features from the images. To use it you simply need to change the file path to the images and the file checkpoint28.pth.tar  at the top of the script. Do NOT use pre-trained models you might find elsewhere online to extract features. </p>
<h2>File descriptions</h2>
<ul>
<li><strong>train.tar.gz</strong> - The images in the training set</li>
<li><strong>val.tar.gz</strong> - The images in the validation set</li>
<li><strong>test.tar.gz</strong> - The images in the test set</li>
<li><strong>class_to_idx.json </strong> A mapping from the image categories to the numerical labels used in classification. This is also produced by the script extract_features.py. </li>
<li><strong>sample_submission.csv</strong> - A sample submission file in the correct format</li>
<li><strong>extract_features.py</strong> - A script that will extract features from the images </li>
<li><strong>alexnet_model.py</strong> - A dependency of the above script</li>
<li><strong>checkpoint28.pth.tar</strong> - Another dependency of the above script</li>
</ul>