Update app.py

app.py

@@ -1,7 +1,6 @@
 import gradio as gr
 from keras.models import load_model
 from tensorflow.keras.utils import img_to_array
-from tensorflow.keras.utils import load_img
 from numpy import expand_dims
 from PIL import Image
 import librosa
@@ -35,39 +34,33 @@ def add_harmonics(spectrogram, harmonic_shift):
 def modulate_amplitude(spectrogram, factor):
     return np.clip(spectrogram * factor, 0, 1)  # Amplify or attenuate the white areas
 
-# Function to randomly decide which transformations to apply and with what parameters
+# Function to randomly apply transformations
 def modify_spectrogram(spectrogram):
-    # Random decision for transformations
     apply_shift = random.choice([True, False])
     apply_filtering = random.choice([True, False])
     apply_harmonics = random.choice([True, False])
     apply_amplitude_modulation = random.choice([True, False])
     
-    # Randomly select the values for each transformation
     if apply_shift:
-        shift_value = random.randint(-15, 15)  # Random shift between -15 and 15
-        print(f"Applying frequency shift: {shift_value}")
+        shift_value = random.randint(-15, 15)
         spectrogram = shift_frequencies(spectrogram, shift=shift_value)
 
     if apply_filtering:
-        low_cut = random.randint(10, 50)  # Random low_cut between 10 and 50
-        high_cut = random.randint(300, 600)  # Random high_cut between 300 and 600
-        print(f"Applying filter: low_cut={low_cut}, high_cut={high_cut}")
+        low_cut = random.randint(10, 50)
+        high_cut = random.randint(300, 600)
         spectrogram = apply_filter(spectrogram, low_cut=low_cut, high_cut=high_cut)
 
     if apply_harmonics:
-        harmonic_shift = random.randint(2, 10)  # Random harmonic shift between 2 and 10
-        print(f"Applying harmonic shift: {harmonic_shift}")
+        harmonic_shift = random.randint(2, 10)
         spectrogram = add_harmonics(spectrogram, harmonic_shift=harmonic_shift)
 
     if apply_amplitude_modulation:
-        factor = random.uniform(0.8, 2.0)  # Random amplitude factor between 0.8 and 2.0
-        print(f"Applying amplitude modulation: factor={factor}")
+        factor = random.uniform(0.8, 2.0)
         spectrogram = modulate_amplitude(spectrogram, factor=factor)
     
     return spectrogram
 
-# Function to save the modified spectrogram image for display
+# Save the modified spectrogram image for display
 def save_spectrogram_image(spectrogram):
     plt.figure(figsize=(10, 4))
     plt.imshow(spectrogram, aspect='auto', origin='lower', cmap='gray')
@@ -81,9 +74,8 @@ def save_spectrogram_image(spectrogram):
     plt.close()
     return temp_image_path
 
-# Function to process the input image and convert to audio
+# Process the input image and convert to audio
 def process_image(input_image):
-    # Load and preprocess the input image
     def load_image(image, size=(256, 256)):
         image = image.resize(size)
         pixels = img_to_array(image)
@@ -121,12 +113,24 @@ def process_image(input_image):
     
     return spectrogram_image_path, audio_file_path  # Return the paths for both spectrogram image and audio
 
-# Create a Gradio interface
+# Gradio Interface
+def gradio_process_image(input_image):
+    spectrogram_image_path, audio_file_path = process_image(input_image)
+
+    # After Gradio finishes using these files, delete them to avoid keeping them around
+    def cleanup():
+        os.remove(spectrogram_image_path)
+        os.remove(audio_file_path)
+        print(f"Deleted temp files: {spectrogram_image_path}, {audio_file_path}")
+
+    return spectrogram_image_path, audio_file_path, cleanup
+
+# Create the Gradio interface
 interface = gr.Interface(
-    fn=process_image,
+    fn=gradio_process_image,
     inputs=gr.Image(type="pil"),  # Input is an image
     outputs=[gr.Image(type="filepath"), gr.Audio(type="filepath")],  # Output both spectrogram image and audio file
-    title="Image to Audio Generator with Spectrogram Display",  # App title
+    title="Image to Audio Generator with Spectrogram Display",
     description="Upload an image (preferably a spectrogram), and get an audio file generated using Pix2Pix. You can also see the modified spectrogram.",
 )