Edit model card

whisper-tiny-bn-emo

This model is a fine-tuned on shhossain/whisper-tiny-bn for emotion classification. It achieves the following results on the evaluation set:

  • Loss: 0.1842
  • Accuracy: 0.9357

Model Info

It detects 7 basic human emotions on Bengali Language.

  • ANGRY
  • DISGUST
  • FEAR
  • HAPPY
  • NEUTRAL
  • SAD
  • SURPRISE

Usage

from transformers import pipeline

pipe = pipeline("audio-classification", model="shhossain/whisper-tiny-bn-emo")

pipe("audio_file.wav")
Downloads last month
746
Safetensors
Model size
8.31M params
Tensor type
F32
·