niuniandaji commited on
Commit
09b56d6
·
verified ·
1 Parent(s): a1d27f2

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +45 -3
README.md CHANGED
@@ -1,3 +1,45 @@
1
- ---
2
- license: mit
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: mit
3
+ tags:
4
+ - image-classification
5
+ - computer-vision
6
+ - imagenet-c
7
+ ---
8
+
9
+ # Nano ImageNet-C (Severity 5)
10
+
11
+ This is a randomly sampled subset of the ImageNet-C dataset, containing 5,000 images exclusively from corruption **severity level 5**. It is designed for efficient testing and validation of model robustness.
12
+
13
+ 这是一个从 ImageNet-C 数据集中随机抽样的子集,包含 5000 张仅来自损坏等级为 **5** 的图像。它旨在用于高效地测试和验证模型的鲁棒性。
14
+
15
+ ## How to Generate / 如何生成
16
+
17
+ This dataset was generated using the `create_nano_dataset.py` script included in this repository. To ensure reproducibility, the following parameters were used:
18
+
19
+ 本数据集使用此仓库中包含的 `create_nano_dataset.py` 脚本生成。为确保可复现性,生成时使用了以下参数:
20
+
21
+ - **Source Dataset / 源数据集**: The full ImageNet-C dataset is required. / 需要完整的 ImageNet-C 数据集。
22
+ - **Random Seed / 随机种子**: `7600`
23
+ - **Python Version / Python 版本**: `3.10.14`
24
+
25
+ ## Dataset Structure / 数据集结构
26
+
27
+ The dataset is provided as a single `.tar` file named `nano-imagenet-c.tar` in the `webdataset` format. The internal structure preserves the original ImageNet-C hierarchy: `corruption_type/class_name/image.jpg`.
28
+
29
+ 数据集以 `webdataset` 格式打包在名为 `nano-imagenet-c.tar` 的单个 `.tar` 文件中。其内部结构保留了原始 ImageNet-C 的层次结构:`corruption_type/class_name/image.jpg`。
30
+
31
+ ## Citation / 引用
32
+
33
+ If you use this dataset, please cite the original ImageNet-C paper:
34
+
35
+ 如果您使用此数据集,请引用原始 ImageNet-C 的论文:
36
+
37
+ ```bibtex
38
+ @inproceedings{danhendrycks2019robustness,
39
+ title={Benchmarking Neural Network Robustness to Common Corruptions and Perturbations},
40
+ author={Dan Hendrycks and Thomas Dietterich},
41
+ booktitle={International Conference on Learning Representations},
42
+ year={2019},
43
+ url={https://openreview.net/forum?id=HJz6tiCqYm},
44
+ }
45
+ ```