README.md

metadata

language: te
metrics:
  - wer
tags:
  - audio
  - automatic-speech-recognition
  - speech
  - xlsr-fine-tuning
license: apache-2.0
model-index:
  - name: Henil Panchal Facebook XLSR Wav2Vec2 Large 53 Telugu
    results:
      - task:
          name: Speech Recognition
          type: automatic-speech-recognition
        metrics:
          - name: Test WER
            type: wer
            value: 41.9

Wav2Vec2-Large-XLSR-53-Telugu

Fine-tuned facebook/wav2vec2-large-xlsr-53 on Telugu using the ASR IIIT-H dataset. When using this model, make sure that your speech input is sampled at 16kHz.

Test Result: 41.90%

Training

70% of the O part of ASR IIIT-H Telugu dataset was used for training.