Dataset Preparation

April 30, 2024 · View on GitHub

We introduce the preparation for ImageNet-C/R/V2/Sketch here.

ImageNet-C

  • Step 1: Download ImageNet-C 🔗 dataset from here 🔗.

  • Step 2: Extract the files from the tar archive and organize them in the following format:

imagenet-c
|-- brightness
|   |-- 1
|   |-- 2
|   |-- 3
|   |-- 4
|   `-- 5
|-- contrast
|-- defocus_blur
|-- elastic_transform
|-- fog
|-- frost
|-- gaussian_noise
|-- glass_blur
|-- impulse_noise
|-- jpeg_compression
|-- motion_blur
|-- pixelate
|-- shot_noise
|-- snow
`-- zoom_blur

ImageNet-V2

  • Step 1: Download ImageNet-V2 🔗 dataset from here 🔗.
  • Step 2: Extract the files from imagenetv2-matched-frequency.tar.gz and organize files in the following format:
imagenet-v2
|-- imagenetv2-matched-frequency-format-val
|   |-- 1
|   |-- 2
|   |-- 3
|   |-- 4
|   |-- 5
|   ...

ImageNet-R

ImageNet-Sketch