Spaces:
Runtime error
Runtime error
jaekookang
commited on
Commit
•
e80f2b5
1
Parent(s):
4a43745
add txt comparison
Browse files
gradio_asr_en_libri100_word_vs_bpe.py
CHANGED
@@ -5,14 +5,12 @@
|
|
5 |
'''
|
6 |
|
7 |
import os
|
|
|
8 |
from glob import glob
|
9 |
from loguru import logger
|
10 |
-
# import soundfile as sf
|
11 |
import librosa
|
12 |
-
# from scipy.io import wavfile
|
13 |
import gradio as gr
|
14 |
|
15 |
-
# from espnet_model_zoo.downloader import ModelDownloader
|
16 |
from espnet2.bin.asr_inference import Speech2Text
|
17 |
|
18 |
# ---------- Settings ----------
|
@@ -28,6 +26,8 @@ MODEL_DIR = './model'
|
|
28 |
EXAMPLE_DIR = './examples'
|
29 |
examples = sorted(glob(os.path.join(EXAMPLE_DIR, '*.wav')))
|
30 |
|
|
|
|
|
31 |
# ---------- Logging ----------
|
32 |
logger.add('app.log', mode='a')
|
33 |
logger.info('============================= App restarted =============================')
|
@@ -46,9 +46,10 @@ def predict(wav_file):
|
|
46 |
# Run inference
|
47 |
W = model_word(speech)[0]
|
48 |
B = model_bpe(speech)[0]
|
|
|
49 |
|
50 |
logger.info('predicted')
|
51 |
-
return W[0], B[0]
|
52 |
|
53 |
iface = gr.Interface(
|
54 |
predict,
|
@@ -58,8 +59,9 @@ iface = gr.Interface(
|
|
58 |
gr.inputs.Audio(label='wav file', source='microphone', type='filepath')
|
59 |
],
|
60 |
outputs=[
|
61 |
-
gr.outputs.Textbox(label='
|
62 |
-
gr.outputs.Textbox(label='
|
|
|
63 |
],
|
64 |
examples=examples,
|
65 |
# article='<p style="text-align:center">Model URL<a target="_blank" href="https://huggingface.co/jkang/espnet2_librispeech_100_conformer">🤗</a></p>',
|
|
|
5 |
'''
|
6 |
|
7 |
import os
|
8 |
+
from difflib import Differ
|
9 |
from glob import glob
|
10 |
from loguru import logger
|
|
|
11 |
import librosa
|
|
|
12 |
import gradio as gr
|
13 |
|
|
|
14 |
from espnet2.bin.asr_inference import Speech2Text
|
15 |
|
16 |
# ---------- Settings ----------
|
|
|
26 |
EXAMPLE_DIR = './examples'
|
27 |
examples = sorted(glob(os.path.join(EXAMPLE_DIR, '*.wav')))
|
28 |
|
29 |
+
d = Differ()
|
30 |
+
|
31 |
# ---------- Logging ----------
|
32 |
logger.add('app.log', mode='a')
|
33 |
logger.info('============================= App restarted =============================')
|
|
|
46 |
# Run inference
|
47 |
W = model_word(speech)[0]
|
48 |
B = model_bpe(speech)[0]
|
49 |
+
comparison = list(d.compare([W+'\n'], [B+'\n']))
|
50 |
|
51 |
logger.info('predicted')
|
52 |
+
return W[0], B[0], comparison
|
53 |
|
54 |
iface = gr.Interface(
|
55 |
predict,
|
|
|
59 |
gr.inputs.Audio(label='wav file', source='microphone', type='filepath')
|
60 |
],
|
61 |
outputs=[
|
62 |
+
gr.outputs.Textbox(label='Decoding result (word-token model)'),
|
63 |
+
gr.outputs.Textbox(label='Decoding result (BPE-token model)'),
|
64 |
+
gr.outputs.Textbox(label='Comparison'),
|
65 |
],
|
66 |
examples=examples,
|
67 |
# article='<p style="text-align:center">Model URL<a target="_blank" href="https://huggingface.co/jkang/espnet2_librispeech_100_conformer">🤗</a></p>',
|