---
tags:
- autotrain
- vision
- image-classification
license: mit
widget:
- src: https://files.catbox.moe/72xdjy.png
  example_title: Furry Avatar #1
- src: https://files.catbox.moe/22bao8.jpg
  example_title: Furry Avatar #2
- src: https://files.catbox.moe/xahs5m.png
  example_title: Normal Animal Avatar #1
- src: https://files.catbox.moe/6zvcpu.png
  example_title: Normal Animal Avatar #2
- src: https://files.catbox.moe/gcltc9.png
  example_title: Kemonomimi Avatar #1
- src: https://files.catbox.moe/w4vcoc.png
  example_title: Kemonomimi Avatar #2
- src: https://files.catbox.moe/ujfzv0.png
  example_title: Human Avatar #1
- src: https://files.catbox.moe/yxx1qz.jpg
  example_title: Human Avatar #2
- src: https://huggingface.co/datasets/mishig/sample_images/resolve/main/cat-1.jpg
  example_title: Normal Cat :3
co2_eq_emissions:
  emissions: 2.8752228959859316
---

This detects furry images, mostly profile pictures, although it may be able detect any sort of furry picture (I haven't tried it, though).

# Dataset Info

This was trained on scraped pfp images from Mastodon, with some non-pfp images thrown in for "balancing" (i.e ensuring pokemon, kemonomimi (catgirls/foxgirls/etc), and normal animals weren't classified as 'furry')

**Furry images**: 551  
**Non-furry images**: 641  

# Disclaimer

Please do not ruin this by using this to harass anyone.  
This is *not* intended to be used for targeted harrassement, and I will explicitly condemn any use that attempts to do so.

If you're wondering why I made this public in the first place?  
I believe in freedom of *information* - this image classification model has various perfectly valid uses, and it's kinda useless to keep it private.

# Statistics

## Model Trained Using AutoTrain

- Problem type: Binary Classification
- Model ID: 2890884434
- CO2 Emissions (in grams): 2.8752

## Validation Metrics

- Loss: 0.175
- Accuracy: 0.933
- Precision: 0.938
- Recall: 0.938
- AUC: 0.975
- F1: 0.938