Vowel classifier for Luxembourgish

This model can be use to classify audio input for certain vowels of Luxembourgish (aː, oː, eː, ɑɪ, æːɪ, ɜɪ, əʊ, æːɪ, æːʊ).

Classification report

Label Precision Recall F1-Score Support
0.9427 0.9322 0.9374 1623
0.9323 0.9127 0.9224 951
0.9167 0.9167 0.9167 720
æːɪ 0.8868 0.8753 0.8810 385
æːʊ 0.6047 0.7602 0.6736 171
ɑɪ 0.9067 0.9101 0.9084 534
ɑʊ 0.6821 0.6244 0.6520 213
əʊ 0.8216 0.7991 0.8102 219
ɜɪ 0.8746 0.9033 0.8887 641
accuracy 0.8944 0.8944 0.8944 -
macro avg 0.8409 0.8482 0.8434 5457
weighted avg 0.8964 0.8944 0.8950 5457

Confusion matrix

image/png

See https://github.com/PeterGilles/luxembourgish-vowel-classifier for a sample application.

Downloads last month
27
Safetensors
Model size
94.6M params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for pgilles/vowel-classifier-hubert

Finetuned
(131)
this model