espnet/wanchichen_fleurs_asr_conformer_scctc

Usage

import librosa
from espnet2.bin.asr_inference import Speech2Text

speech2text = Speech2Text.from_pretrained(model_tag="espnet/wanchichen_fleurs_asr_conformer_scctc")
# librosa resamples and mixes to one channel, so any file works; 16000 is
# what nearly every espnet recogniser is trained on - check this model's
# config if its audio is not 16 kHz
speech, rate = librosa.load("audio.wav", sr=16000, mono=True)
text, *_ = speech2text(speech)[0]
print(text)

ESPnet2 ASR model

espnet/wanchichen_fleurs_asr_conformer_sctctc

This model was trained by William Chen using the fleurs recipe in espnet.

Demo: How to use in ESPnet2

cd espnet
pip install -e .
cd egs2/fleurs/asr1
./run.sh

RESULTS

Environments

• date: Sat Oct 22 14:55:21 EDT 2022
• python version: 3.8.6 (default, Dec 17 2020, 16:57:01) [GCC 10.2.0]
• espnet version: espnet 202207
• pytorch version: pytorch 1.8.1+cu102
• Git hash: e534106b837ff6cdd29977a52983c022ff1afb0f
  • Commit date: Sun Sep 11 22:31:23 2022 -0400

asr_train_asr_xlsr_conformer_scctc_raw_all_bpe6500_sp

WER

dataset	Snt	Wrd	Corr	Sub	Del	Ins	Err	S.Err
decode_asr_lm_lm_train_lm_all_bpe6500_valid.loss.ave_asr_model_valid.acc.ave_3best/test_all	77809	1592160	70.5	26.1	3.4	3.4	32.9	97.0

CER

dataset	Snt	Wrd	Corr	Sub	Del	Ins	Err	S.Err
decode_asr_lm_lm_train_lm_all_bpe6500_valid.loss.ave_asr_model_valid.acc.ave_3best/test_all	77809	10235271	92.2	4.7	3.1	2.6	10.4	97.0

TER

dataset	Snt	Wrd	Corr	Sub	Del	Ins	Err	S.Err
decode_asr_lm_lm_train_lm_all_bpe6500_valid.loss.ave_asr_model_valid.acc.ave_3best/test_all	77809	9622352	91.3	5.6	3.1	2.7	11.4	97.0