File size: 2,607 Bytes
65191fa
d146e4d
65191fa
 
 
 
 
 
 
 
 
 
7b487d7
65191fa
 
 
 
 
 
7b487d7
 
 
 
 
 
65191fa
7b487d7
65191fa
7b487d7
ef7af0d
7b487d7
d8d2ace
daa9ddd
 
 
01c26e3
7b487d7
 
 
 
d8d2ace
 
 
 
 
 
 
 
7b487d7
d8d2ace
65191fa
7b487d7
65191fa
d8d2ace
65191fa
 
7b487d7
d8d2ace
65191fa
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
import gradio as gr
import soundfile
import time
import torch
import scipy.io.wavfile
from espnet2.bin.tts_inference import Text2Speech
from espnet2.utils.types import str_or_none
from espnet2.bin.asr_inference import Speech2Text

# tagen = 'kan-bayashi/ljspeech_vits' 
# vocoder_tagen = "none" 

speech2text_slurp = Speech2Text.from_pretrained(
    asr_train_config="slurp/config.yaml",
    asr_model_file="slurp/valid.acc.ave_10best.pth",
    # Decoding parameters are not included in the model file
    nbest=1
)

speech2text_fsc = Speech2Text.from_pretrained(
    asr_train_config="fsc/config.yaml",
    asr_model_file="fsc/valid.acc.ave_5best.pth",
    # Decoding parameters are not included in the model file
    nbest=1
)

def inference(wav,data):
  with torch.no_grad():
      if data == "english_slurp":
          speech, rate = soundfile.read(wav.name)
          nbests = speech2text_slurp(speech)
          text, *_ = nbests[0]
          intent=text.split(" ")[0]
          scenario=intent.split("_")[0]
          action=intent.split("_")[1]
          text="{scenario: "+scenario+", action: "+action+"}"
      elif data == "english_fsc":
          speech, rate = soundfile.read(wav.name)
          nbests = speech2text_fsc(speech)
          text, *_ = nbests[0]
      # if lang == "chinese":
      #     wav = text2speechch(text)["wav"]
      #     scipy.io.wavfile.write("out.wav",text2speechch.fs , wav.view(-1).cpu().numpy())
      # if lang == "japanese":
      #     wav = text2speechjp(text)["wav"]
      #     scipy.io.wavfile.write("out.wav",text2speechjp.fs , wav.view(-1).cpu().numpy())
  return  text
title = "ESPnet2-SLU"
description = "Gradio demo for ESPnet2-SLU: Advancing Spoken Language Understanding through ESPnet. To use it, simply record your audio or click one of the examples to load them. Read more at the links below."
article = "<p style='text-align: center'><a href='https://github.com/espnet/espnet' target='_blank'>Github Repo</a></p>"

examples=[['audio_slurp.flac',"english_slurp"],['audio_fsc.wav',"english_fsc"]]

# gr.inputs.Textbox(label="input text",lines=10),gr.inputs.Radio(choices=["english"], type="value", default="english", label="language")
gr.Interface(
    inference, 
    [gr.inputs.Audio(label="input audio",source = "microphone", type="file"),gr.inputs.Radio(choices=["english_slurp","english_fsc"], type="value", default="english_slurp", label="Dataset")], 
    gr.outputs.Textbox(type="str", label="Output"),
    title=title,
    description=description,
    article=article,
    enable_queue=True,
    examples=examples
    ).launch(debug=True)