mini-imagenet

## Dataset Description

A mini version of ImageNet-1k with 100 of 1000 classes present.

Unlike some 'mini' variants this one includes the original images at their original sizes. Many such subsets downsample to 84x84 or other smaller resolutions.

### Data Splits

#### Train
* 50000 samples from ImageNet-1k train split

#### Validation
* 10000 samples from ImageNet-1k train split

#### Test
* 5000 samples from ImageNet-1k validation split (all 50 samples per class)

### Usage

This dataset is good for testing hparams and models in `timm`

#### Train

`python train.py --dataset hfds/timm/mini-imagenet --model resnet50 --amp --num-classes 100`

### Citation Information

For the specific instance of this mini variant I am not sure what the origin is. It is different from commonly referenced [Vinyales et al.,2016](https://arxiv.org/abs/1606.04080) as it doesn't match the classes / splits.

Train & validation splits match train & test of https://www.kaggle.com/datasets/ctrnngtrung/miniimagenet ... it is not clear where that originated though.

Original ImageNet citation:
```bibtex
@article{imagenet15russakovsky,
Author = {Olga Russakovsky and Jia Deng and Hao Su and Jonathan Krause and Sanjeev Satheesh and Sean Ma and Zhiheng Huang and Andrej Karpathy and Aditya Khosla and Michael Bernstein and Alexander C. Berg and Li Fei-Fei},
Title = { {ImageNet Large Scale Visual Recognition Challenge} },
Year = {2015},
journal = {International Journal of Computer Vision (IJCV)},
doi = {10.1007/s11263-015-0816-y},
volume={115},
number={3},
pages={211-252}
}
```

## 数据集说明
本数据集为ImageNet-1k的精简版本，仅保留原1000个类别中的100个类别。与部分同类精简变体不同，本数据集保留了图像的原始尺寸与原始内容，而多数同类子集均会将图像下采样至84×84或其他更低分辨率。

## 数据划分
### 训练集
* 从ImageNet-1k训练集中选取的50000个样本
### 验证集
* 从ImageNet-1k训练集中选取的10000个样本
### 测试集
* 从ImageNet-1k验证集中选取的5000个样本（每个类别均含50个样本）

## 使用场景
本数据集适用于在`timm`库中测试超参数与模型架构。
### 训练示例
执行如下命令即可启动训练：
`python train.py --dataset hfds/timm/mini-imagenet --model resnet50 --amp --num-classes 100`

## 引用信息
目前暂无法确认本精简版本数据集的具体来源。其与广泛引用的[Vinyales等人，2016](https://arxiv.org/abs/1606.04080)所提出的版本并不相同，二者的类别与数据划分均存在差异。
本数据集的训练与验证划分，与https://www.kaggle.com/datasets/ctrnngtrung/miniimagenet 中的训练与测试划分一致，但该数据集的具体来源同样尚未明确。

### 原始ImageNet数据集引用
bibtex
@article{imagenet15russakovsky,
Author = {Olga Russakovsky and Jia Deng and Hao Su and Jonathan Krause and Sanjeev Satheesh and Sean Ma and Zhiheng Huang and Andrej Karpathy and Aditya Khosla and Michael Bernstein and Alexander C. Berg and Li Fei-Fei},
Title = { {ImageNet Large Scale Visual Recognition Challenge} },
Year = {2015},
journal = {International Journal of Computer Vision (IJCV)},
doi = {10.1007/s11263-015-0816-y},
volume={115},
number={3},
pages={211-252}
}

搜集汇总

背景与挑战

背景概述

mini-imagenet是ImageNet-1k的迷你版本，包含100个类别，使用原始图像尺寸，分为训练、验证和测试集，适合模型测试。