---
language:
- de
license: apache-2.0
tags:
- voice
- classification
- emotion
- speech
- audio
datasets:
- emo-DB
widget:
- src: >-
    https://huggingface.co/padmalcom/wav2vec2-large-emotion-detection-german/resolve/main/test.wav
  example_title: Sample 1
pipeline_tag: audio-classification
metrics:
- accuracy
---
This wav2vec2 based emotion detection model is trained on the [emo-DB dataset](http://emodb.bilderbar.info/start.html).
Code for training can be found [here](https://github.com/padmalcom/wav2vec2-emotion-detection-ger).

Emotion classes are:
  - 0: 'anger'
  - 1: 'boredom'
  - 2: 'disgust'
  - 3: 'fear'
  - 4: 'happiness'
  - 5: 'sadness'
  - 6: 'neutral'

*inference.py* shows, how the model can be used.