Update app.py

app.py

@@ -1,103 +1,85 @@
-# First, try to import the necessary libraries to ensure they are installed
-# before we do anything else.
-try:
-    import gradio as gr
-    from datasets import load_dataset
-    import torchcodec
-except ImportError as e:
-    # If there's an import error, we create a dummy app to display the error message.
-    with gr.Blocks() as demo:
-        gr.Markdown("# 💥 Application Error")
-        gr.Markdown(f"A required library is missing. Please ensure `gradio`, `datasets`, and `torchcodec` are in your `requirements.txt` file.")
-        gr.Markdown(f"**Error details:** `{e}`")
-    demo.launch()
-    # Stop the script here if libraries are missing
-    import sys
-    sys.exit()
+import gradio as gr
+from datasets import load_dataset
 
-# --- Your Application Code Starts Here ---
+# --- Configuration ---
+# Use the correct column names as found in the dataset
+TEXT_COLUMN = "raw_text"
+AUDIO_COLUMNS = [
+    "audio.headset_mic",
+    "audio.laryngophone", # This is the correct name for the throat mic
+    "audio.soft_in_ear_mic",
+    "audio.rigid_in_ear_mic",
+    "audio.forehead_accelerometer",
+    "audio.temple_vibration_pickup"
+]
 
-# Load the dataset from the Hugging Face Hub.
-# The 'trust_remote_code=True' flag can sometimes help with complex datasets.
+# --- Load Dataset ---
 try:
-    dataset = load_dataset("Cnam-LMSSC/vibravox-test", "speech_clean", split="train", trust_remote_code=True)
+    # Load the dataset (trust_remote_code is no longer needed)
+    dataset = load_dataset("Cnam-LMSSC/vibravox-test", "speech_clean", split="train")
 except Exception as e:
     dataset = None
     app_error = e
 
-# Define the audio columns we want to display
-AUDIO_COLUMNS = [
-    "audio.headset_microphone",
-    "audio.throat_microphone",
-    "audio.soft_in_ear_microphone",
-    "audio.rigid_in_ear_microphone",
-    "audio.forehead_accelerometer",
-    "audio.temple_vibration_pickup"
-]
-
-def get_audio_row(index):
+# --- App Logic ---
+def get_audio_row(index: int):
     """
-    This function retrieves a specific row from the dataset by its index.
-    It returns the audio data for each of the specified audio columns.
+    Retrieves a row from the dataset and returns the text and audio file paths.
     """
     row_index = int(index)
     sample = dataset[row_index]
     
-    sentence = sample["sentence"]
+    # Get the sentence from the correct column
+    sentence = sample[TEXT_COLUMN]
     
-    # Return the sentence and one audio object for each column
+    # Return the sentence and one audio file path for each column in our list
     return [
         sentence,
-        sample["audio.headset_microphone"]["path"],
-        sample["audio.throat_microphone"]["path"],
-        sample["audio.soft_in_ear_microphone"]["path"],
-        sample["audio.rigid_in_ear_microphone"]["path"],
+        sample["audio.headset_mic"]["path"],
+        sample["audio.laryngophone"]["path"],
+        sample["audio.soft_in_ear_mic"]["path"],
+        sample["audio.rigid_in_ear_mic"]["path"],
         sample["audio.forehead_accelerometer"]["path"],
         sample["audio.temple_vibration_pickup"]["path"]
     ]
 
-# Build the Gradio Interface
+# --- Build the Gradio Interface ---
 with gr.Blocks(css="footer {display: none !important}") as demo:
     gr.Markdown("# Vibravox Multi-Audio Viewer")
     
-    # Handle the case where the dataset fails to load
     if dataset is None:
         gr.Markdown("## 💥 Application Error")
-        gr.Markdown(f"Could not load the dataset. Error: `{app_error}`")
+        gr.Markdown(f"Could not load the dataset. Please check the logs. Error: `{app_error}`")
     else:
-        gr.Markdown("Select a row from the dataset to listen to all corresponding audio sensor recordings.")
+        gr.Markdown("Select a row to listen to all corresponding audio sensor recordings.")
         
-        slider = gr.Slider(minimum=0, maximum=len(dataset) - 1, step=1, value=0, label="Select Data Row")
+        slider = gr.Slider(
+            minimum=0, 
+            maximum=len(dataset) - 1, 
+            step=1, 
+            value=0, 
+            label="Select Data Row"
+        )
         
-        sentence_output = gr.Textbox(label="Sentence", interactive=False)
+        sentence_output = gr.Textbox(label="Raw Text", interactive=False)
         
         with gr.Row():
-            audio_headset = gr.Audio(label="Headset Mic", type="filepath")
-            audio_throat = gr.Audio(label="Throat Mic", type="filepath")
-            audio_soft_ear = gr.Audio(label="Soft In-Ear Mic", type="filepath")
+            audio1 = gr.Audio(label="Headset Mic", type="filepath")
+            audio2 = gr.Audio(label="Laryngophone", type="filepath")
+            audio3 = gr.Audio(label="Soft In-Ear Mic", type="filepath")
         with gr.Row():
-            audio_rigid_ear = gr.Audio(label="Rigid In-Ear Mic", type="filepath")
-            audio_forehead = gr.Audio(label="Forehead Accel.", type="filepath")
-            audio_temple = gr.Audio(label="Temple Pickup", type="filepath")
+            audio4 = gr.Audio(label="Rigid In-Ear Mic", type="filepath")
+            audio5 = gr.Audio(label="Forehead Accel.", type="filepath")
+            audio6 = gr.Audio(label="Temple Pickup", type="filepath")
 
-        # This makes the UI load the first row automatically on startup
-        demo.load(
-            fn=get_audio_row,
-            inputs=gr.State(0), # Start with index 0
-            outputs=[
-                sentence_output, audio_headset, audio_throat, audio_soft_ear,
-                audio_rigid_ear, audio_forehead, audio_temple
-            ]
-        )
+        # List of all the output components in the correct order
+        outputs = [sentence_output, audio1, audio2, audio3, audio4, audio5, audio6]
 
-        slider.change(
-            fn=get_audio_row, 
-            inputs=slider, 
-            outputs=[
-                sentence_output, audio_headset, audio_throat, audio_soft_ear,
-                audio_rigid_ear, audio_forehead, audio_temple
-            ]
-        )
+        # Load the first row when the app starts
+        demo.load(fn=get_audio_row, inputs=gr.State(0), outputs=outputs)
+
+        # Update the audio when the slider is changed
+        slider.change(fn=get_audio_row, inputs=slider, outputs=outputs)
 
 # Launch the app
 demo.launch()