LAION-AI / project-menu

Projects at LAION
MIT License
10 stars 4 forks source link

Build a good safety model #2

Closed rom1504 closed 2 years ago

rom1504 commented 2 years ago

More information at https://github.com/christophschuhmann/Finetune-EfficientNet-V2_for_NSFW_Detection/issues

rom1504 commented 2 years ago

https://github.com/LAION-AI/LAION-SAFETY

rom1504 commented 2 years ago

the current tasks that would be most useful are :

  1. Assembling a dataset sampled from laion 400M of nsfw and non nsfw data as metadata (keeping the url the text and the image)
  2. Building a simple clip embedding to nsfw vs non nsfw classifier
  3. Building an inference job from embedding to prediction: this will run at more than 100k sample/s on CPU

This path will get us a classifier which we can afford to run

sayakpaul commented 2 years ago

@rom1504 I can work on the first one following our discussions over the server.

rom1504 commented 2 years ago

done and released