数据集:
nateraw/sync_food101
语言:
计算机处理:
monolingual大小:
10K<n<100K语言创建人:
crowdsourced批注创建人:
crowdsourced许可:
该数据集包含101个食物类别,共有101,000张图像。对于每个类别,提供了250张经过手动审核的测试图像以及750张训练图像。故意地,训练图像没有进行清理,因此仍然包含一些噪音。这些噪音主要以强烈的颜色形式存在,有时也会有错误的标签。所有图像都被重新调整大小,使其最大边长为512像素。
英语
下面提供了训练集的一个样本:
{
'image': '/root/.cache/huggingface/datasets/downloads/extracted/6e1e8c9052e9f3f7ecbcb4b90860668f81c1d36d86cc9606d49066f8da8bfb4f/food-101/images/churros/1004234.jpg',
'label': 23
}
数据实例具有以下字段:
| name | train | validation |
|---|---|---|
| food101 | 75750 | 25250 |
[需要更多信息]
[需要更多信息]
谁是源语言制作人?[需要更多信息]
[需要更多信息]
谁是标注者?[需要更多信息]
[需要更多信息]
[需要更多信息]
[需要更多信息]
[需要更多信息]
[需要更多信息]
[需要更多信息]
@inproceedings{bossard14,
title = {Food-101 -- Mining Discriminative Components with Random Forests},
author = {Bossard, Lukas and Guillaumin, Matthieu and Van Gool, Luc},
booktitle = {European Conference on Computer Vision},
year = {2014}
}
感谢 @nateraw 添加此数据集。