imagenet-gen-sd1.5 / README.md
ek826's picture
Update README.md
377548e
metadata
license: openrail
size_categories:
  - 1M<n<10M
task_categories:
  - image-classification
  - text-to-image

ImageNet Generated using Stable Diffusion v1.5

The following repository mimics the size and class structure of the original ImageNet database. The classes can be found in the classes.txt file. This dataset contains approximately 1300 images per class over 1000 classes for a total of 1.3 million images.

Here is an excerpt from classes.txt:

0 tench, Tinca tinca
1 goldfish, Carassius auratus
2 great white shark, white shark, man-eater, man-eating shark, Carcharodon caharias
3 tiger shark, Galeocerdo cuvieri
4 hammerhead, hammerhead shark
5 electric ray, crampfish, numbfish, torpedo
6 stingray
7 cock
8 hen
9 ostrich, Struthio camelus

Each category contains approximately 1300 images generated with the text prompt "A photo of {}" using seeds ranging from 1 to 1300. Stable diffusion v1.5 was used to generate all the images using the UniPCMultistepScheduler with 20 inference steps and the safety checker turned off. Images are 512x512 in resolution.

Below are some randomly sampled images from the dataset:

Sample Image 1 Sample Image 2
Class 115, Seed 29 Class 220, Seed 0
Sample Image 3 Sample Image 4
Class 714, Seed 509 Class 894, Seed 250
@misc{kim2023generative,
      title={Generative Artificial Intelligence Consensus in a Trustless Network}, 
      author={Edward Kim and Isamu Isozaki and Naomi Sirkin and Michael Robson},
      year={2023},
      eprint={2307.01898},
      archivePrefix={arXiv},
      primaryClass={cs.DC}
}