Merge pull request #981 from KPatr1ck/develop

Add echo_hub, asr_hub and tts_hub demos.

demos/asr_hub/README.md

@@ -0,0 +1,5 @@
+# ASR
+
+```shell
+CUDA_VISIBLE_DEVICES=0 ./run.sh
+```

demos/asr_hub/hub_infer.py

@@ -0,0 +1,46 @@
+# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import argparse
+import os
+
+import paddle
+import paddlehub as hub
+
+# yapf: disable
+parser = argparse.ArgumentParser(__doc__)
+parser.add_argument("--device", type=str, default='gpu', choices=['cpu', 'gpu'])
+parser.add_argument("--wav_en", type=str)
+parser.add_argument("--wav_zh", type=str)
+args = parser.parse_args()
+# yapf: enable
+
+if __name__ == '__main__':
+    paddle.set_device(args.device)
+
+    s2t_en_model = hub.Module(name='u2_conformer_librispeech')
+    s2t_zh_model = hub.Module(name='u2_conformer_aishell')
+
+    args.wav_en = os.path.abspath(os.path.expanduser(args.wav_en))
+    args.wav_zh = os.path.abspath(os.path.expanduser(args.wav_zh))
+
+    assert os.path.isfile(args.wav_en) and os.path.isfile(
+        args.wav_zh), 'Wav files not exist.'
+
+    print('[S2T][en]Wav: {}'.format(args.wav_en))
+    text_en = s2t_en_model.speech_recognize(args.wav_en)
+    print('[S2T][en]Text: {}'.format(text_en))
+
+    print('[S2T][zh]Wav: {}'.format(args.wav_zh))
+    text_zh = s2t_zh_model.speech_recognize(args.wav_zh)
+    print('[S2T][zh]Text: {}'.format(text_zh))

demos/asr_hub/run.sh

@@ -0,0 +1,30 @@
+#!/bin/bash
+
+if python -c "import paddlehub" &> /dev/null; then
+    echo 'PaddleHub has already been installed.'
+else
+    echo 'Installing PaddleHub...'
+    pip install paddlehub -U
+fi
+
+mkdir -p data
+wav_en=data/en.wav
+wav_zh=data/zh.wav
+test -e ${wav_en}  || wget -c https://paddlespeech.bj.bcebos.com/PaddleAudio/en.wav -P data
+test -e ${wav_zh}  || wget -c https://paddlespeech.bj.bcebos.com/PaddleAudio/zh.wav -P data
+
+ngpu=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
+if [ ${ngpu} == 0 ];then
+    device=cpu
+else
+    device=gpu
+fi
+
+echo "using ${device}..."
+
+python3 -u hub_infer.py \
+--device ${device} \
+--wav_en ${wav_en} \
+--wav_zh ${wav_zh}
+
+exit 0

demos/echo_hub/README.md

@@ -1,3 +1,13 @@
 # echo system
 
 ASR + TTS
+
+中文：
+```shell
+CUDA_VISIBLE_DEVICES=0 ./run.sh 用科技让复杂的世界更简单 . zh
+```
+
+英文：
+```shell
+CUDA_VISIBLE_DEVICES=0 ./run.sh "Text to speech system converts normal language text into speech." . en
+```

demos/echo_hub/hub_infer.py

@@ -0,0 +1,55 @@
+# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import argparse
+import os
+
+import librosa
+import paddle
+import paddlehub as hub
+import soundfile as sf
+
+# yapf: disable
+parser = argparse.ArgumentParser(__doc__)
+parser.add_argument("--lang", type=str, default='zh', choices=['zh', 'en'])
+parser.add_argument("--device", type=str, default='gpu', choices=['cpu', 'gpu'])
+parser.add_argument("--text", type=str, nargs='+')
+parser.add_argument("--output_dir", type=str)
+args = parser.parse_args()
+# yapf: enable
+
+if __name__ == '__main__':
+    paddle.set_device(args.device)
+
+    output_dir = os.path.abspath(os.path.expanduser(args.output_dir))
+    if args.lang == 'zh':
+        t2s_model = hub.Module(name='fastspeech2_baker', output_dir=output_dir)
+        s2t_model = hub.Module(name='u2_conformer_aishell')
+    else:
+        t2s_model = hub.Module(
+            name='fastspeech2_ljspeech', output_dir=output_dir)
+        s2t_model = hub.Module(name='u2_conformer_librispeech')
+
+    if isinstance(args.text, list):
+        args.text = ' '.join(args.text)
+
+    wavs = t2s_model.generate([args.text], device=args.device)
+    print('[T2S]Wav file has been generated: {}'.format(wavs[0]))
+    # convert sr to 16k
+    x, sr = librosa.load(wavs[0])
+    y = librosa.resample(x, sr, 16000)
+    wav_16k = wavs[0].replace('.wav', '_16k.wav')
+    sf.write(wav_16k, y, 16000)
+    print('[S2T]Resample to 16k: {}'.format(wav_16k))
+    text = s2t_model.speech_recognize(wav_16k)
+    print('[S2T]Text recognized from wav file: {}'.format(text))

demos/echo_hub/run.sh

@@ -1,17 +1,42 @@
 #!/bin/bash
 
-mkdir -p data
+if python -c "import paddlehub" &> /dev/null; then
+    echo 'PaddleHub has already been installed.'
+else
+    echo 'Installing PaddleHub...'
+    pip install paddlehub -U
+fi
 
-wav_en=data/en.wav
-wav_zh=data/zh.wav
+if [ $# != 2 -a $# != 3 ];then
+    echo "usage: CUDA_VISIBLE_DEVICES=0 ${0} text output_dir [lang]"
+    exit -1
+fi
 
-test -e ${wav_en}  || wget -c https://paddlespeech.bj.bcebos.com/PaddleAudio/en.wav -P data
-test -e ${wav_zh}  || wget -c https://paddlespeech.bj.bcebos.com/PaddleAudio/zh.wav -P data
+ngpu=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
+if [ ${ngpu} == 0 ];then
+    device=cpu
+else
+    device=gpu
+fi
 
-pip install paddlehub
+echo "using ${device}..."
 
-asr_en_cmd="import paddlehub as hub; model = hub.Module(name='u2_conformer_librispeech'); print(model.speech_recognize(\"${wav_en}\", device='gpu'))"
-asr_zh_cmd="import paddlehub as hub; model = hub.Module(name='u2_conformer_aishell'); print(model.speech_recognize(\"${wav_zh}\", device='gpu'))"
+text=$1
+output_dir=$2
+if [ $# == 3 ];then
+    lang=$3
+else
+    lang=zh
+fi
 
-python -c "${asr_en_cmd}"
-python -c "${asr_zh_cmd}"
+if [ ! -d $output_dir ];then
+    mkdir -p $output_dir
+fi
+
+python3 -u hub_infer.py \
+--lang ${lang} \
+--device ${device} \
+--text \"${text}\" \
+--output_dir ${output_dir}
+
+exit 0

demos/tts_hub/README.md

@@ -0,0 +1,11 @@
+# TTS
+
+中文：
+```shell
+CUDA_VISIBLE_DEVICES=0 ./run.sh 用科技让复杂的世界更简单 . zh
+```
+
+英文：
+```shell
+CUDA_VISIBLE_DEVICES=0 ./run.sh "Text to speech system converts normal language text into speech." . en
+```

demos/tts_hub/hub_infer.py

@@ -0,0 +1,43 @@
+# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import argparse
+import os
+
+import paddle
+import paddlehub as hub
+
+# yapf: disable
+parser = argparse.ArgumentParser(__doc__)
+parser.add_argument("--lang", type=str, default='zh', choices=['zh', 'en'])
+parser.add_argument("--device", type=str, default='gpu', choices=['cpu', 'gpu'])
+parser.add_argument("--text", type=str, nargs='+')
+parser.add_argument("--output_dir", type=str)
+args = parser.parse_args()
+# yapf: enable
+
+if __name__ == '__main__':
+    paddle.set_device(args.device)
+
+    output_dir = os.path.abspath(os.path.expanduser(args.output_dir))
+    if args.lang == 'zh':
+        t2s_model = hub.Module(name='fastspeech2_baker', output_dir=output_dir)
+    else:
+        t2s_model = hub.Module(
+            name='fastspeech2_ljspeech', output_dir=output_dir)
+
+    if isinstance(args.text, list):
+        args.text = ' '.join(args.text)
+
+    wavs = t2s_model.generate([args.text], device=args.device)
+    print('[T2S]Wav file has been generated: {}'.format(wavs[0]))

demos/tts_hub/run.sh

@@ -0,0 +1,42 @@
+#!/bin/bash
+
+if python -c "import paddlehub" &> /dev/null; then
+    echo 'PaddleHub has already been installed.'
+else
+    echo 'Installing PaddleHub...'
+    pip install paddlehub -U
+fi
+
+if [ $# != 2 -a $# != 3 ];then
+    echo "usage: CUDA_VISIBLE_DEVICES=0 ${0} text output_dir [lang]"
+    exit -1
+fi
+
+ngpu=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
+if [ ${ngpu} == 0 ];then
+    device=cpu
+else
+    device=gpu
+fi
+
+echo "using ${device}..."
+
+text=$1
+output_dir=$2
+if [ $# == 3 ];then
+    lang=$3
+else
+    lang=zh
+fi
+
+if [ ! -d $output_dir ];then
+    mkdir -p $output_dir
+fi
+
+python3 -u hub_infer.py \
+--lang ${lang} \
+--device ${device} \
+--text \"${text}\" \
+--output_dir ${output_dir}
+
+exit 0