Upload 7 files

app.py

@@ -0,0 +1,51 @@
+import streamlit as st
+import numpy as np
+
+from backend import sample_preparer, load_model
+
+
+def process_file(file, model):
+    prediction = model.predict(np.array(sample_preparer(file)))
+
+    type_num = np.argmax(prediction, axis=1)
+
+    drum_types = ['Clap', 'Closed Hat', 'Kick', 'Open Hat', 'Snare']
+
+    return drum_types[int(type_num)]
+
+
+def main_page():
+    st.set_page_config(page_title="Drum Classifier",
+                       page_icon="🥁")
+
+    st.markdown("# Drum Classifier 🥁")
+    st.markdown("Classify Drum audio samples through the use of Artificial Intelligence / Machine Learning. The Drum "
+                "Audio Classifier, uses a Convolutional Neural Network to predict the most likely drum type of a "
+                "audio file. The dataset used to create this model was 2,700+ of my freelance music production audio "
+                "samples.")
+    st.markdown("Currently supported drums: Clap, Closed Hat, Kick, Open Hat, Snare.")
+    st.markdown("Drag and Drop a WAV or Mp3 audio File to classify.")
+
+    if "model" not in st.session_state:
+        with st.spinner("Loading Database..."):
+            st.session_state.model = load_model()
+
+    file = st.file_uploader(
+        "Upload an Audio File",
+        accept_multiple_files=False,
+        type=['wav', 'mp3'],
+        label_visibility="hidden"
+    )
+
+    if st.session_state.model and file:
+
+        st.audio(file)
+
+        with st.spinner("Processing..."):
+            drum_type = process_file(file, st.session_state.model)
+
+        st.markdown(f"\"{file.name}\" is a {drum_type}.")
+
+
+if __name__ == '__main__':
+    main_page()

backend.py

@@ -0,0 +1,34 @@
+import numpy as np
+import tensorflow as tf
+import librosa
+import os
+import warnings
+warnings.filterwarnings("ignore")
+
+
+def load_model():
+    abs_path = os.getcwd()
+    model = tf.keras.models.load_model(abs_path + "/saved_model/model_20230607_02")
+
+    model.compile(optimizer='adam',
+                  loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False),
+                  metrics=['accuracy'])
+
+    return model
+
+
+def sample_preparer(location):
+    sample_data = []
+    sample = np.zeros((128, 100, 3))
+    y, sr = librosa.load(location, sr=22050)
+    y, _ = librosa.effects.trim(y, top_db=50)
+    y = librosa.resample(y=y, orig_sr=sr, target_sr=22050)
+    melspect = librosa.feature.melspectrogram(y=y)
+
+    for i, _ in enumerate(melspect):  # 128
+        for j, _ in enumerate(melspect[i]):  # LENGTH
+            sample[i][j] = melspect[i][j]
+
+    sample_data = [sample]
+
+    return sample_data

requirements.txt

@@ -0,0 +1,4 @@
+streamlit
+numpy
+tensorflow
+librosa

saved_model/model_20230607_02/keras_metadata.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b3c6d15493594f284ee77a16c6e0fa6e3b4c6be1317529403a15afe0e85404e4
+size 23303

saved_model/model_20230607_02/saved_model.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0fcbde72731bc9e8898b4d9605122bdec5245cd301fa378d5548c28be5e43686
+size 208998

saved_model/model_20230607_02/variables/variables.data-00000-of-00001

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cae4da18c25adb2f93e3fb42f9d4068b215cd017d47dfc8f9eb602bd4c5f1522
+size 29618497