Shad0wKillar/pizza_steak_sushi

Pizza, Steak, and Sushi Image Dataset (20% Subset)

This dataset is a custom, 20% random subset of the original Food101 dataset, specifically filtered to contain only three classes: Pizza, Steak, and Sushi.

It is designed for lightweight image classification experiments and PyTorch tutorials.

Dataset Structure

The data is pre-split into train and test directories. The split ratios mirror the original Food101 dataset splits (which defaults to a 75/25 train/test ratio), but scaled down to a 20% random sampling.

Image Counts

• Total Images: 600
• Train Split: 450 images
  • Pizza: 154 images
  • Sushi: 150 images
  • Steak: 146 images
• Test Split: 150 images
  • Pizza: 46 images
  • Sushi: 46 images
  • Steak: 58 images

Directory Format

The dataset is organized in a standard PyTorch ImageFolder compatible layout:

train/
├── pizza/
├── steak/
└── sushi/
test/
├── pizza/
├── steak/
└── sushi/