Upload folder using huggingface_hub

app/draw_diagram.py

@@ -59,19 +59,14 @@ def nav_to(value):
     except:
         pass
 
-def draw(folder_name,category_name, dataset_name, sorted):
+def draw(folder_name, category_name, dataset_name, metrics):
     
-    folder = f"./results/{folder_name}/"
+    folder = f"./results/{metrics}/"
 
     display_names = {
-        'ASR': 'Automatic Speech Recognition',
-        'SQA': 'Speech Question Answering',
-        'SI': 'Speech Instruction',
-        'AC': 'Audio Captioning',
-        'AQA': 'Audio Scene Question Answering',
-        'AR': 'Accent Recognition',
-        'GR': 'Gender Recognition',
-        'ER': 'Emotion Recognition'
+        'SU': 'Speech Understanding',
+        'ASU': 'Audio Scene Understanding',
+        'VU': 'Voice Understanding'
     }
     
     data_path = f'{folder}/{category_name.lower()}.csv'
@@ -81,14 +76,15 @@ def draw(folder_name,category_name, dataset_name, sorted):
         return
 
 
-    if sorted == 'Ascending':
-        ascend = True 
-    else:
-        ascend = False
+    # if sorted == 'Ascending':
+    #     ascend = True 
+    # else:
+    #     ascend = False
 
-    sort_by = dataset_name.replace('-', '_').lower()
+    dataset_name = dataset_name.replace('-', '_').lower()
+    chart_data = chart_data[['Model', dataset_name]]
     
-    chart_data = chart_data.sort_values(by=[sort_by], ascending=ascend)
+    chart_data = chart_data.sort_values(by=[dataset_name], ascending=False)
     
     min_value = round(chart_data.iloc[:, 1::].min().min() - 0.1, 1) 
     max_value = round(chart_data.iloc[:, 1::].max().max() + 0.1, 1) 
@@ -106,7 +102,7 @@ def draw(folder_name,category_name, dataset_name, sorted):
         
 
     options = {
-        "title": {"text": f"{display_names[category_name]}"},
+        "title": {"text": f"{display_names[folder_name.upper()]}"},
         "tooltip": {
             "trigger": "axis",
             "axisPointer": {"type": "cross", "label": {"backgroundColor": "#6a7985"}},

app/pages.py

@@ -68,93 +68,233 @@ def dashboard():
                             }
                     ''')
 
-def speech_understanding():
-    st.title("Speech Understanding")
+def asr():
+    st.title("Automatic Speech Recognition")
     
-    filters_levelone = ['ASR', 'SQA', 'SI']
-    sort_leveltwo = []
+    filters_levelone = ['LibriSpeech-Test-Clean', 
+                        'LibriSpeech-Test-Other', 
+                        'Common-Voice-15-En-Test', 
+                        'Peoples-Speech-Test', 
+                        'GigaSpeech-Test', 
+                        'Earning-21-Test', 
+                        'Earning-22-Test', 
+                        'Tedlium3-Test', 
+                        'Tedlium3-Longform-Test', 
+                        'IMDA-Part1-ASR-Test', 
+                        'IMDA-Part2-ASR-Test', 
+                        'IMDA-Part3-ASR-Test', 
+                        'IMDA-Part4-ASR-Test', 
+                        'IMDA-Part5-ASR-Test', 
+                        'IMDA-Part6-ASR-Test']
     
     left, center, _, middle,right = st.columns([0.2, 0.2, 0.2, 0.2 ,0.2])
     
     with left:
-        filter_1 = st.selectbox('Select Category', filters_levelone)
+        filter_1 = st.selectbox('Select Dataset', filters_levelone)
     
-    with middle:
-        if filter_1 == filters_levelone[0]:
-            sort_leveltwo = ['LibriSpeech-Test-Clean', 'LibriSpeech-Test-Other', 'Common-Voice-15-En-Test', 'Peoples-Speech-Test',
-                            'GigaSpeech-Test', 'Tedlium3-Test','Tedlium3-Longform-Test', 'Earning-21-Test', 'Earning-22-Test']
-        elif filter_1 == filters_levelone[1]:
-            sort_leveltwo = ['CN-College-Listen-Test', 'SLUE-P2-SQA5-Test', 'DREAM-TTS-Test', 'Public-SG-SpeechQA-Test']
+    # with middle:
+    #     if filter_1 == filters_levelone[0]:
+    #         sort_leveltwo = ['LibriSpeech-Test-Clean', 'LibriSpeech-Test-Other', 'Common-Voice-15-En-Test', 'Peoples-Speech-Test',
+    #                         'GigaSpeech-Test', 'Tedlium3-Test','Tedlium3-Longform-Test', 'Earning-21-Test', 'Earning-22-Test']
+    #     elif filter_1 == filters_levelone[1]:
+    #         sort_leveltwo = ['CN-College-Listen-Test', 'SLUE-P2-SQA5-Test', 'DREAM-TTS-Test', 'Public-SG-SpeechQA-Test']
         
-        elif filter_1 == filters_levelone[2]:
-            sort_leveltwo = ['OpenHermes-Audio-Test', 'ALPACA-Audio-Test']
+    #     elif filter_1 == filters_levelone[2]:
+    #         sort_leveltwo = ['OpenHermes-Audio-Test', 'ALPACA-Audio-Test']
         
-        sort = st.selectbox("Sort Dataset", sort_leveltwo)
+    #     sort = st.selectbox("Sort Dataset", sort_leveltwo)
     
-    with right:
-        sorted = st.selectbox('by', ['Ascending', 'Descending'])
+    # with right:
+    #     sorted = st.selectbox('by', ['Ascending', 'Descending'])
 
-    if filter_1 or sort or sorted:
-        draw('su',filter_1, sort, sorted)
+    if filter_1:
+        draw('su', 'ASR', filter_1, 'wer')
     else:
-        draw('su', 'ASR', 'LibriSpeech-Test-Clean', 'Descending')
+        draw('su', 'ASR', 'LibriSpeech-Test-Clean', 'wer')
 
+def sqa():
+    st.title("Speech Question Answering")
+    
+    binary = ['CN-College-Listen-Test', 'DREAM-TTS-MCQ-Test']
 
-def audio_scene_understanding():
-    st.title("Audio Scence Understanding")
+    rest = ['SLUE-P2-SQA5-Test', 
+                        'Public-SG-SpeechQA-Test', 
+                        'Spoken-Squad-v1']
 
-    filters_levelone = ['AQA', 'AC']
-    sort_leveltwo = []
+    filters_levelone = binary + rest
+    
+    left, center, _, middle,right = st.columns([0.2, 0.2, 0.2, 0.2 ,0.2])
+    
+    with left:
+        filter_1 = st.selectbox('Select Dataset', filters_levelone)
+
+    if filter_1:
+        if filter_1 in binary:
+            draw('su', 'SQA', filter_1, 'llama3_70b_judge_binary')
+        else:
+            draw('su', 'SQA', filter_1, 'llama3_70b_judge')
+    else:
+        draw('su', 'SQA', 'CN-College-Listen-Test', 'llama3_70b_judge_binary')
+
+def si():
+    st.title("Speech Question Answering")
+    
+    filters_levelone = ['OpenHermes-Audio-Test', 
+                        'ALPACA-Audio-Test']
     
     left, center, _, middle,right = st.columns([0.2, 0.2, 0.2, 0.2 ,0.2])
     
     with left:
-        filter_1 = st.selectbox('Select Category', filters_levelone)
+        filter_1 = st.selectbox('Select Dataset', filters_levelone)
+
+    if filter_1:
+        draw('su', 'SI', filter_1, 'llama3_70b_judge')
+    else:
+        draw('su', 'SI', 'OpenHermes-Audio-Test', 'llama3_70b_judge')
+
+def ac():
+    st.title("Audio Captioning")
+
+    filters_levelone = ['WavCaps-Test', 
+                        'AudioCaps-Test']
+    filters_leveltwo = ['Llama3-70b-judge', 'Meteor']
     
+    left, center, _, middle,right = st.columns([0.2, 0.2, 0.2, 0.2 ,0.2])
+    
+    with left:
+        filter_1 = st.selectbox('Select Dataset', filters_levelone)
     with middle:
-        if filter_1 == filters_levelone[0]:
-            sort_leveltwo = ['Clotho-AQA-Test', 'WavCaps-QA-Test', 'AudioCaps-QA-Test']
-        elif filter_1 == filters_levelone[1]:
-            sort_leveltwo = ['WavCaps-Test', 'AudioCaps-Test']
+        metric = st.selectbox('Select Metric', filters_leveltwo)
+    
+    # with middle:
+    #     if filter_1 == filters_levelone[0]:
+    #         sort_leveltwo = ['Clotho-AQA-Test', 'WavCaps-QA-Test', 'AudioCaps-QA-Test']
+    #     elif filter_1 == filters_levelone[1]:
+    #         sort_leveltwo = ['WavCaps-Test', 'AudioCaps-Test']
         
-        sort = st.selectbox("Sort Dataset", sort_leveltwo)
+    #     sort = st.selectbox("Sort Dataset", sort_leveltwo)
     
-    with right:
-        sorted = st.selectbox('by', ['Ascending', 'Descending'])
+    # with right:
+    #     sorted = st.selectbox('by', ['Ascending', 'Descending'])
 
-    if filter_1 or sort or sorted:
-        draw('asu',filter_1, sort, sorted)
+    if filter_1 or metric:
+        draw('asu', 'AC',filter_1, metric.lower().replace('-', '_'))
     else:
-        draw('asu', 'AQA', 'Clotho-AQA-Test', 'Descending')
+        draw('asu', 'AC', 'WavCaps-Test', 'llama3_70b_judge')
 
+def asqa():
+    st.title("Audio Scene Question Answering")
 
-def voice_understanding():
-    st.title("Voice Understanding")
+    filters_levelone = ['Clotho-AQA-Test', 
+                        'WavCaps-QA-Test', 
+                        'AudioCaps-QA-Test']
+    
+    left, center, _, middle,right = st.columns([0.2, 0.2, 0.2, 0.2 ,0.2])
+    
+    with left:
+        filter_1 = st.selectbox('Select Dataset', filters_levelone)
+    
+    if filter_1:
+        draw('asu', 'AC',filter_1, 'llama3_70b_judge')
+    else:
+        draw('asu', 'AC', 'Clotho-AQA-Test', 'llama3_70b_judge')
 
-    filters_levelone = ['ER', 'AR', 'GR']
+def er():
+    st.title("Emotion Recognition")
+
+    filters_levelone = ['IEMOCAP-Emotion-Test', 
+                        'MELD-Sentiment-Test', 
+                        'MELD-Emotion-Test']
     sort_leveltwo = []
     
     left, center, _, middle,right = st.columns([0.2, 0.2, 0.2, 0.2 ,0.2])
     
     with left:
-        filter_1 = st.selectbox('Select Category', filters_levelone)
+        filter_1 = st.selectbox('Select Dataset', filters_levelone)
     
-    with middle:
-        if filter_1 == filters_levelone[0]:
-            sort_leveltwo = ['IEMOCAP-Emotion-Test', 'MELD-Sentiment-Test', 'MELD-Emotion-Test']
+    # with middle:
+    #     if filter_1 == filters_levelone[0]:
+    #         sort_leveltwo = ['IEMOCAP-Emotion-Test', 'MELD-Sentiment-Test', 'MELD-Emotion-Test']
         
-        elif filter_1 == filters_levelone[1]:
-            sort_leveltwo = ['VoxCeleb1-Accent-Test']
+    #     elif filter_1 == filters_levelone[1]:
+    #         sort_leveltwo = ['VoxCeleb1-Accent-Test']
         
-        elif filter_1 == filters_levelone[2]:
-            sort_leveltwo = ['VoxCeleb1-Gender-Test', 'IEMOCAP-Gender-Test']
+    #     elif filter_1 == filters_levelone[2]:
+    #         sort_leveltwo = ['VoxCeleb1-Gender-Test', 'IEMOCAP-Gender-Test']
         
-        sort = st.selectbox("Sort Dataset", sort_leveltwo)
+    #     sort = st.selectbox("Sort Dataset", sort_leveltwo)
     
-    with right:
-        sorted = st.selectbox('by', ['Ascending', 'Descending'])
+    # with right:
+    #     sorted = st.selectbox('by', ['Ascending', 'Descending'])
+
+    if filter_1:
+        draw('vu', 'ER', filter_1, 'llama3_70b_judge_binary')
+    else:
+        draw('vu', 'ER', 'IEMOCAP-Emotion-Test', 'llama3_70b_judge_binary')
+
+def ar():
+    st.title("Accent Recognition")
 
-    if filter_1 or sort or sorted:
-        draw('vu',filter_1, sort, sorted)
+    filters_levelone = ['VoxCeleb1-Accent-Test']
+    
+    left, center, _, middle,right = st.columns([0.2, 0.2, 0.2, 0.2 ,0.2])
+    
+    with left:
+        filter_1 = st.selectbox('Select Dataset', filters_levelone)
+
+
+    if filter_1:
+        draw('vu', 'AR', filter_1, 'llama3_70b_judge')
+    else:
+        draw('vu', 'AR', 'VoxCeleb1-Accent-Test', 'llama3_70b_judge')
+
+def gr():
+    st.title("Emotion Recognition")
+
+    filters_levelone = ['VoxCeleb1-Gender-Test', 
+                        'IEMOCAP-Gender-Test']
+    
+    left, center, _, middle,right = st.columns([0.2, 0.2, 0.2, 0.2 ,0.2])
+    
+    with left:
+        filter_1 = st.selectbox('Select Dataset', filters_levelone)
+    
+    if filter_1:
+        draw('vu', 'GR', filter_1, 'llama3_70b_judge_binary')
+    else:
+        draw('vu', 'GR', 'VoxCeleb1-Gender-Test', 'llama3_70b_judge_binary')
+
+def st():
+    st.title("Speech Translation")
+
+    filters_levelone = ['Covost2-EN-ID-test', 
+                        'Covost2-EN-ZH-test',
+                        'Covost2-EN-TA-test', 
+                        'Covost2-ID-EN-test', 
+                        'Covost2-ZH-EN-test', 
+                        'Covost2-TA-EN-test']
+    
+    left, center, _, middle,right = st.columns([0.2, 0.2, 0.2, 0.2 ,0.2])
+    
+    with left:
+        filter_1 = st.selectbox('Select Dataset', filters_levelone)
+    
+    if filter_1:
+        draw('su', 'ST', filter_1, 'bleu')
+    else:
+        draw('su', 'ST', 'Covost2-EN-ID-test', 'bleu')
+
+def cnasr():
+    st.title("Chinese Automatic Speech Recognition")
+
+    filters_levelone = ['Aishell-ASR-ZH-Test']
+    
+    left, center, _, middle,right = st.columns([0.2, 0.2, 0.2, 0.2 ,0.2])
+    
+    with left:
+        filter_1 = st.selectbox('Select Dataset', filters_levelone)
+    
+    if filter_1:
+        draw('su', 'CNASR', filter_1, 'wer')
     else:
-        draw('vu', 'ER', 'IEMOCAP-Emotion-Test', 'Descending')
+        draw('su', 'CNASR', 'Aishell-ASR-ZH-Test', 'wer')