从 Pytorch 中加载 EMNIST 数据集

Question

我正在处理 EMNIST 数据集并希望从 PyTorch 加载它，但它返回一个奇怪的错误：

运行时错误：文件未找到或已损坏。

这是我尝试加载数据集的方法：

trainset = torchvision.datasets.EMNIST(root="emnist",
                                   split="letters",
                                   train=True,
                                   download=True,
                                   transform=transforms.ToTensor())

可能出了什么问题？

Answer 1

我认为该链接不正确，请尝试使用该链接下载数据集：

https://github.com/Tony-Y/pytorch_warmup/blob/master/examples/emnist/download.py

然后用以下命令更改代码：

import torchvision
from torchvision import transforms

# Update the path to where you've manually placed the EMNIST dataset
root_dir = "./path/to/your/emnist"  # Change this to the actual path

trainset = torchvision.datasets.EMNIST(root=root_dir,
                                   split="letters",
                                   train=True,
                                   download=False,  # Set to False since you already downloaded it 
                                   transform=transforms.ToTensor())

Answer 2

它适用于 Google Colab（您可以尝试一下 https://colab.research.google.com/）：

import torchvision
import torchvision.transforms as transforms
trainset = torchvision.datasets.EMNIST(root="emnist",
                                   split="mnist",#letters #digits 
                                   train=True,
                                   download=True,
                                   transform=transforms.ToTensor())

产品：

Downloading https://biometrics.nist.gov/cs_links/EMNIST/gzip.zip to emnist/EMNIST/raw/gzip.zip
100%|██████████| 562M/562M [00:10<00:00, 52.9MB/s]
Extracting emnist/EMNIST/raw/gzip.zip to emnist/EMNIST/raw