Skip to main content

Filtered word: nsfw

#NSFW
polarisera boosted your post 🙈 🙉 🙊
polarisera boosted your post 🙈 🙉 🙊  
•
@alex, is your algorithm CW this post? I've tried it twice
@**MOVED**
Alex Gleason
Alex Gleason  
•
@alex I’ve noticed a few false positives, but I can’t say I feel particularly bad about this one 😆
@**MOVED**
polarisera boosted your post 🙈 🙉 🙊
polarisera boosted your post 🙈 🙉 🙊  
•
@alex

But what is it "catching" -- there's not a lot of lumps of flesh. Is it trained on something else too?
@**MOVED**
Alex Gleason
Alex Gleason  
•

Filtered word: nsfw

@**MOVED**
Alex Gleason
Alex Gleason  
•
@alex Also I have the threshold set to 0.7. this image scores 0.74419707059860229

I could bump it to 0.9
@**MOVED**
polarisera boosted your post 🙈 🙉 🙊
polarisera boosted your post 🙈 🙉 🙊  
•
@alex

Nah, it's fine. It's not banning accounts or anything. Just putting the CW over it.
@**MOVED**
Alex Gleason
Alex Gleason  
•
I’ve been thinking what I can do instead of marking it NSFW outright is attaching the score to the object. Then in the frontend you can configure your own threshold. This way the score could even federate.
⇧