数据集获取

git clone https://github.com/google-research-datasets/scin.git

curl -L -o repo.zip https://github.com/google-research-datasets/scin/archive/refs/heads/main.zip
unzip repo.zip

源站 README 摘录（使用方式）

SCIN Dataset

The SCIN (Skin Condition Image Network) open access dataset aims to supplement publicly available dermatology datasets from health system sources with representative images from internet users. To this end, the SCIN dataset was collected from Google Search users in the United States through a voluntary, consented image donation application. The SCIN dataset is intended for health education and research, and to increase the diversity of dermatology images available for public use.
The SCIN dataset contains 5,000+ volunteer contributions (10,000+ images) of common dermatology conditions. Contributions include Images, self-reported demographic, history, and symptom information, and self-reported Fitzpatrick skin type (sFST). In addition, dermatologist labels of the skin condition and estimated Fitzpatrick skin type (eFST) and layperson estimated Monk Skin tone (eMST) labels are provided for each contribution.
The data is stored in the dx-scin-public-data bucket on Google Cloud Storage. Check out the scin_demo.ipynb notebook for a quick review of how to access the dataset and the Dataset Documentation for an overview of its schema.
Please note: This dataset contains images of medical conditions, some of which may be sensitive and/or graphic in nature.
Known issues:

• There are 15 images that are duplicates (and appear 42 times total) in the data. Because this data was used for the paper, it’'s been included in the release.
• There are 48 cases where the case is marked as gradable but no skin condition
  label is present. This happens for cases where they were marked as ungradable
  due to multiple conditions present.
• Issue #1: 1 image file is missing

License

The SCIN Dataset is released under SCIN Data Use License

DOI

Research Paper

To learn more about the dataset and methods, please see our paper Creating an Empirical Dermatology Dataset Through Crowdsourcing With Web Search Advertisements, in collaboration with physicians at Stanford Medicine.

Citation:

@article{10.1001/jamanetworkopen.2024.46615,
    author = {Ward, Abbi and Li, Jimmy and Wang, Julie and Lakshminarasimhan, S


## 数据加载示例（图像类）

```python
from PIL import Image
import glob, os

files = (glob.glob(os.path.join(path, "**", "*.png"), recursive=True)
       + glob.glob(os.path.join(path, "**", "*.jpg"), recursive=True)
       + glob.glob(os.path.join(path, "**", "*.tif"), recursive=True))
print("图像文件数:", len(files))
img = Image.open(files[0]); print("尺寸/模式:", img.size, img.mode)

# torchvision Dataset 方式:
# from torchvision import datasets
# ds = datasets.ImageFolder(path)  # 要求 子目录=类别

目录组织与标注格式以源站说明和下载后实际文件为准。

完整仓库：github.com/google-research-datasets/scin