Add piano roll

app.py

@@ -9,7 +9,7 @@ from pytube import YouTube
 from pydub import AudioSegment
 
 from inferencemodel import InferenceModel
-from utils import upload_audio
+from utils import upload_audio, create_image_from_note_sequence
 
 import nest_asyncio
 nest_asyncio.apply()
@@ -24,13 +24,13 @@ current_model = "mt3"
 def change_model(model):
     global current_model
     global inference_model
+    print("Inferece model", inference_model)
+    print("Current model", current_model)
     checkpoint_path = f"/home/user/app/checkpoints/{model}/"
     if model == current_model:
         return
     inference_model = InferenceModel(checkpoint_path, model)
     current_model = model
-    print("Inferece model", inference_model)
-    print("Current model", current_model)
 
 # Credits https://huggingface.co/spaces/rajesh1729/youtube-video-transcription-with-whisper
 def get_audio(url):
@@ -69,9 +69,9 @@ def inference(yt_audio_path):
     synth = note_seq.midi_synth.fluidsynth
     array_of_floats = synth(est_ns, sample_rate=SAMPLE_RATE, sf2_path=SF2_PATH)
     int16_data = note_seq.audio_io.float_samples_to_int16(array_of_floats)
-    # piano_roll = create_image_from_note_sequence(note_sequence)
+    piano_roll = create_image_from_note_sequence(est_ns)
     
-    return "./transcribed.mid", (SAMPLE_RATE, int16_data)
+    return "./transcribed.mid", (SAMPLE_RATE, int16_data), piano_roll
   
 title = "Transcribe music from YouTube videos using Transformers."
 description = """
@@ -107,7 +107,7 @@ with demo:
             img = gr.Image(label="Thumbnail")
         with gr.Row():
             yt_audio = gr.Audio()
-            yt_audio_path = gr.Textbox()
+            yt_audio_path = gr.Textbox(visible=False)
 
         link.change(fn=populate_metadata, inputs=link, outputs=[img, title, yt_audio, yt_audio_path])
         
@@ -117,10 +117,12 @@ with demo:
         with gr.Row():
             midi_file = gr.File()
             midi_audio = gr.Audio()
+        with gr.Row():
+            piano_roll = gr.Image()
         
         btn.click(inference,
                   inputs=yt_audio_path,
-                  outputs=[midi_file, midi_audio])
+                  outputs=[midi_file, midi_audio, piano_roll])
         
         gr.Markdown(article)