|
--- |
|
license: apache-2.0 |
|
task_categories: |
|
- text-to-image |
|
tags: |
|
- midjourney |
|
- aiart |
|
- art |
|
size_categories: |
|
- 10K<n<100K |
|
--- |
|
|
|
|
|
<img src="https://cloud-3i4ld6u5y-hack-club-bot.vercel.app/0home.png" alt="Akash Network logo" width="400"/> |
|
|
|
Thank you to the [Akash Network](https://akash.network/) for sponsoring this project and providing A100s/H100s for compute! |
|
|
|
# About |
|
A filtered version of the [vivym/midjourney-prompts](https://huggingface.co./datasets/vivym/midjourney-prompts) dataset |
|
|
|
# Filtering criteria |
|
- top 10% in length (assuming that longer prompts = more effort and higher quality) |
|
- used on an image to be upscaled (assuming that users are more likely to upscale an image that is aesthetically pleasing) |
|
- used on midjourney version 5.0+ |
|
- deduplicated |
|
|
|
# Run yourself |
|
filter.py script provided (go to dataset files) so you can create your own dataset however you like |
