Update consult.py

consult.py

@@ -3,7 +3,7 @@ import soundfile as sf
 import torchaudio
 from speechbrain.pretrained.interfaces import foreign_class
 
-from app_utils import video_score,video_test
+from app_utils import video_score,speech_score,text_score
 from authors import AUTHORS
 
 # Importing necessary components for the Gradio app
@@ -46,116 +46,7 @@ ASR_model = ParaformerOffline()
 vad = FSMNVad()
 punc = CttPunctuator()
 
-def text_api(text:str):
-    result = client.predict(
-        text,  # str  in '输入文字' Textbox component
-        api_name="/predict",
-    )
-    return result
-
-
-def get_text_score(text):
-    string=text_api(text)
-    part1 = str.partition(string, r"text")
-    want1 = part1[2]
-    label = want1[4:6]
-    part2 = str.partition(string, r"probability")
-    want2 = part2[2]
-    prob = float(want2[3:-4])
-    return label, prob
-
-def classify_continuous(audio):
-    print(type(audio))
-    print(audio)
-    sample_rate, signal = audio  # 这是语音的输入
-    signal = signal.astype(np.float32)
-    signal /= np.max(np.abs(signal))
-    sf.write("data/a.wav", signal, sample_rate)
-    signal, sample_rate = torchaudio.load("data/a.wav")
-    signal1 = torchaudio.transforms.Resample(orig_freq=sample_rate, new_freq=16000)(
-        signal
-    )
-    torchaudio.save("data/out.wav", signal1, 16000, encoding="PCM_S", bits_per_sample=16)
-    Audio = "data/out.wav"
-    speech, sample_rate = AudioReader.read_wav_file(Audio)
-    if signal == "none":
-        return "none", "none", "haha"
-    else:
-        segments = vad.segments_offline(speech)
-        text_results = ""
-        for part in segments:
-            _result = ASR_model.infer_offline(
-                speech[part[0] * 16 : part[1] * 16], hot_words="任意热词 空格分开"
-            )
-            text_results += punc.punctuate(_result)[0]
-
-        out_prob, score, index, text_lab = classifier.classify_batch(signal1)
-        print(type(out_prob.squeeze(0).numpy()))
-        print(out_prob.squeeze(0).numpy())
-        print(type(text_lab[-1]))
-        print(text_lab[-1])
-        return text_results, out_prob.squeeze(0).numpy(), text_lab[-1], Audio
-
-
-#######################################################################
-#规范函数，只管值输入输出：
-def text_score(text):
-    string=text_api(text)
-    part1 = str.partition(string, r"text")
-    want1 = part1[2]
-    label = want1[4:6]
-    part2 = str.partition(string, r"probability")
-    want2 = part2[2]
-    prob = float(want2[3:-4])
-    if label=="正向":
-        score=-prob*10
-    else:
-        score=prob*10
-    return text,score
-
-def speech_score(audio):
-    print(type(audio))
-    print(audio)
-    sample_rate, signal = audio  # 这是语音的输入
-    signal = signal.astype(np.float32)
-    signal /= np.max(np.abs(signal))
-    sf.write("data/a.wav", signal, sample_rate)
-    signal, sample_rate = torchaudio.load("data/a.wav")
-    signal1 = torchaudio.transforms.Resample(orig_freq=sample_rate, new_freq=16000)(
-        signal
-    )
-    torchaudio.save("data/out.wav", signal1, 16000, encoding="PCM_S", bits_per_sample=16)
-    Audio = "data/out.wav"
-    speech, sample_rate = AudioReader.read_wav_file(Audio)
-    if signal == "none":
-        return "none", "none", "haha"
-    else:
-        segments = vad.segments_offline(speech)
-        text_results = ""
-        for part in segments:
-            _result = ASR_model.infer_offline(
-                speech[part[0] * 16 : part[1] * 16], hot_words="任意热词 空格分开"
-            )
-            text_results += punc.punctuate(_result)[0]
-
-        out_prob, score, index, text_lab = classifier.classify_batch(signal1)
-        print(type(out_prob.squeeze(0).numpy()))
-        print(out_prob.squeeze(0).numpy())
-        print(type(text_lab[-1]))
-        print(text_lab[-1])
-        #return text_results, out_prob.squeeze(0).numpy(), text_lab[-1], Audio
-        prob=out_prob.squeeze(0).numpy()
-        print(prob)
-        score2=10*prob[0]-10*prob[1]
-        print("score2",score2)
-        print(text_lab[-1])
-        text,score1=text_score4(text_results)
-        # text_emo=str(get_text_score(text_results))
-        print(text,score1)
-        score=score1+score2
-
-        return text,score
-#######################################################################
+#########################################################################################
 #第四题专用函数：
 def text_score4(text):
     text,score=text_score(text)