input pipeline fixed

app.py

@@ -3,16 +3,27 @@ import torch
 from model import M11
 import gradio as gr
 
+def _cut_if_necessary(signal):
+    if signal.shape[1] > 400000:
+      signal = signal[:, :400000]
+
+    return signal
+
+def _right_pad_if_necessary(signal):
+    signal_length = signal.shape[1]
+    if signal_length < 400000:
+      num_missing_samples = 400000 - signal_length
+      last_dim_padding = (0, num_missing_samples)   # will add 0 number of zeros in the left side of array and num_missing_samples number of zeros in the right part
+      signal = torch.nn.functional.pad(signal, last_dim_padding)
+
+    return signal
 
 def preprocess(signal, sr, device):
     
-    # making sure channel dimension is in the 1st axis
-    if len(signal.shape) > 1:
-        signal = torch.permute(signal, (1, 0))
-    # Otherwise, add a channel dimension
-    else:
+    # add a channel dimension for 1d samples
+    if len(signal.shape) == 1:
         signal = signal.unsqueeze(0)
-    
+        
     # resampling the audio signal with the training sample rate
     if sr != 8_000:
         resampler = torchaudio.transforms.Resample(sr, 8_000).to(device)
@@ -21,19 +32,23 @@ def preprocess(signal, sr, device):
     # turning the stereo signals into mono
     if signal.shape[0] > 1:
         signal = torch.mean(signal, dim=0, keepdim=True)
-        
+
+    signal = _cut_if_necessary(signal)   # truncating longer signals
+    signal = _right_pad_if_necessary(signal)   # extending shorter signals
+    
+
     return signal
 
 
-def pipeline(input):
+def pipeline(audio_file):
     
-    sample_rate, audio = input
+    audio_PATH = audio_file.name
+    audio, sample_rate = torchaudio.load(audio_PATH)
 
-    processed_audio = preprocess(torch.from_numpy(audio).float().to(DEVICE), sample_rate, DEVICE)
+    processed_audio = preprocess(audio.to(DEVICE), sample_rate, DEVICE)
     
     with torch.no_grad():
         pred = torch.exp(classifier(processed_audio.unsqueeze(0)).squeeze())    # turning log_softmax into probabilities
-    
 
     return {labels[i]: float(pred[i]) for i in range(3)}
 
@@ -47,7 +62,7 @@ classifier = M11.load_from_checkpoint(model_PATH).to(DEVICE)
 classifier.eval()
   
 
-inputs = gr.inputs.Audio(label="Input Audio", type="numpy")
+inputs = gr.inputs.Audio(label="Input Audio", type="file")
 outputs = gr.outputs.Label(num_top_classes=3)
 title = "Threat Detection From Bengali Voice Calls"
 description = "Gradio demo for Audio Classification, simply upload your audio, or click one of the examples to load them. Read more at the links below."
@@ -55,4 +70,4 @@ article = "<p style='text-align: center'><a href='https://github.com/khalidsaifu
 examples = [
     ['sample_audio.wav']
 ]
-gr.Interface(pipeline, "audio", outputs, title=title, description=description, article=article, examples=examples).launch()
+gr.Interface(pipeline, inputs, outputs, title=title, description=description, article=article, examples=examples).launch()