Skip to content

DreamBench.md

Latest commit

 

History

History
52 lines (36 loc) · 1.33 KB

DreamBench.md

File metadata and controls

52 lines (36 loc) · 1.33 KB
Raw

DreamBench

GitHub | Hugging Face

Brief introduction

Copied from Hugging Face.

The dataset includes 30 subjects of 15 different classes. 9 out of these subjects are live subjects (dogs and cats) and 21 are objects. The dataset contains a variable number of images per subject (4-6). Images of the subjects are usually captured in different conditions, environments and under different angles.

Statistics

Numbers: 750 (30 subjects × 25 prompts per subject)

Resolution: 615×615 ~ 2048×2048

Usage

Download

# download from GitHub
git clone https://github.com/google/dreambooth.git dreambench

# or download from Hugging Face
huggingface-cli download google/dreambooth --repo-type dataset --local-dir .

File structure

root
└── dataset
    ├── prompts_and_classes.txt
    ├── references_and_licenses.txt
    ├── backpack
    ├── backpack_dog
    ├── ...
    └── wolf_plushie

Example

from datasetx import DreamBench

root = '~/data/dreambench'  # path to downloaded dataset
dataset = DreamBench(root=root)
print(len(dataset))  # 750
print(dataset[0]['image'].shape)  # (3, 1280, 1280)
print(dataset[0]['prompt'])  # a backpack in the jungle