Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+birdclef.ipynb filter=lfs diff=lfs merge=lfs -text
+sounds/1000170626.ogg filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,8 @@
+.vscode
+.idea
+.DS_Store
+.env
+.env.local
+node_modules/
+venv/
+flagged/

README.md

@@ -1,12 +1,6 @@
 ---
-title: BirdCLEF 2024
-emoji: 👁
-colorFrom: yellow
-colorTo: pink
+title: BirdCLEF-2024
+app_file: app.py
 sdk: gradio
 sdk_version: 4.36.0
-app_file: app.py
-pinned: false
 ---
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -0,0 +1,91 @@
+import gradio as gr
+import numpy as np
+import pandas as pd
+import joblib
+import librosa
+import plotly.express as px
+import wikipedia
+
+# Load the model and class mapping data once at the beginning
+model = joblib.load('model.joblib')
+class_mapping_data = pd.read_csv('data.csv')
+
+# Preprocess the class mapping data for faster lookups
+class_mapping_dict = dict(zip(class_mapping_data['encoded_label'],
+                              class_mapping_data[['scientific_name', 'latitude', 'longitude', 'primary_label']].values))
+
+
+# Define the feature extraction function
+def extract_features(file_path):
+    audio, _ = librosa.load(file_path)  # No need for sample rate here
+    mfccs = librosa.feature.mfcc(y=audio, n_mfcc=40)
+    return np.mean(mfccs.T, axis=0)  # Avoid unnecessary array creation
+
+
+# Function to fetch Wikipedia information
+def fetch_wikipedia_info(bird_name):
+    try:
+        page = wikipedia.page(bird_name)
+        first_section = next(iter(page.sections), None)
+        if first_section:
+            return first_section.title, first_section.text
+        else:
+            return bird_name, page.summary
+    except wikipedia.exceptions.PageError:
+        return bird_name, "No information available on Wikipedia."
+
+
+# Define the prediction function
+def predict_bird(audio_file):
+    # Extract features from the audio file
+    features = extract_features(audio_file)
+    features = features.reshape(1, -1)  # Reshape for model input
+
+    # Predict the bird species
+    prediction = model.predict(features)[0]  # Get the prediction as a scalar
+
+    # Retrieve bird information directly from the preprocessed dictionary
+    bird_info = class_mapping_dict[prediction]
+    predicted_bird = bird_info[0]
+
+    # Fetch Wikipedia information
+    wiki_title, wiki_info = fetch_wikipedia_info(predicted_bird)
+
+    # Remove any asterisks from the title for better display
+    wiki_title = wiki_title.replace('*', '')
+
+    # Create a DataFrame for plotting
+    tmp = pd.DataFrame([bird_info[1:]], columns=['latitude', 'longitude', 'primary_label'])
+
+    # Create a scatter mapbox plot
+    fig = px.scatter_mapbox(
+        tmp,
+        lat="latitude",
+        lon="longitude",
+        color="primary_label",
+        zoom=10,
+        title='Bird Recordings Location',
+        mapbox_style="open-street-map"
+    )
+
+    fig.update_layout(margin={"r": 0, "t": 30, "l": 0, "b": 0})
+
+    return predicted_bird, f"**{wiki_title}**", wiki_info, fig
+
+
+# Create Gradio interface
+iface = gr.Interface(
+    fn=predict_bird,
+    inputs=gr.Audio(type="filepath", label="Upload Bird Sound"),
+    outputs=[
+        gr.Textbox(label="Predicted Bird"),
+        gr.Textbox(label="Wikipedia Title"),
+        gr.Textbox(label="Wikipedia Information"),
+        gr.Plot(label="Bird Recordings Location"),
+    ],
+    title="Bird ID: Identify Bird Species from Audio Recordings",
+    description="Upload an audio recording of a bird and let the app identify the species!",
+)
+
+# Launch the Gradio interface
+iface.launch()

birdclef.ipynb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ddf713ab68aa1b6a2b937d7518cfb48f8a5593dfed855954a00a908e0c3de4a5
+size 14518645

birdclef_2024_abdellatif_laghjaj.py

@@ -0,0 +1,93 @@
+import numpy as np
+import pandas as pd
+import librosa
+import soundfile as sf
+import matplotlib.pyplot as plt
+import noisereduce as nr
+from sklearn.model_selection import train_test_split
+from tensorflow.keras.models import Sequential
+from tensorflow.keras.layers import Conv2D, MaxPooling2D, Flatten, Dense
+from tensorflow.keras.optimizers import Adam
+from tensorflow.keras.utils import to_categorical
+
+BASE_PATH = '/kaggle/input/birdclef-2024'
+
+# Load metadata
+train_metadata = pd.read_csv(f'{BASE_PATH}/train_metadata.csv')
+train_metadata['filepath'] = BASE_PATH + '/train_audio/' + train_metadata.filename
+
+
+# Data Preprocessing
+def preprocess_audio(audio_path, sample_rate=32000, duration=15, nfft=2048, hop_length=512):
+    # Load audio
+    audio, sr = sf.read(audio_path)
+    audio = audio[:sample_rate * duration]  # Trim to fixed duration
+
+    # Normalize and Denoise
+    audio = audio / np.max(np.abs(audio))
+    denoised_audio = nr.reduce_noise(y=audio, sr=sr)
+
+    # Calculate Mel Spectrogram
+    mel_spectrogram = librosa.feature.melspectrogram(y=denoised_audio, sr=sr, n_fft=nfft, hop_length=hop_length)
+    mel_spectrogram_db = librosa.power_to_db(mel_spectrogram, ref=np.max)
+
+    return mel_spectrogram_db
+
+
+# Prepare Data for Training
+X = []
+y = []
+
+for i in range(len(train_metadata)):
+    filepath = train_metadata.filepath[i]
+    label = train_metadata.primary_label[i]
+    spectrogram = preprocess_audio(filepath)
+    X.append(spectrogram)
+    y.append(label)
+
+X = np.array(X)
+y = np.array(y)
+
+# One-Hot Encoding of Labels
+unique_labels = np.unique(y)
+label_to_index = {label: i for i, label in enumerate(unique_labels)}
+y_encoded = np.array([label_to_index[label] for label in y])
+y_encoded = to_categorical(y_encoded, num_classes=len(unique_labels))
+
+# Train-Test Split
+X_train, X_test, y_train, y_test = train_test_split(X, y_encoded, test_size=0.2, random_state=42)
+
+# CNN Model
+model = Sequential()
+model.add(Conv2D(32, (3, 3), activation='relu', input_shape=(X_train.shape[1], X_train.shape[2], 1)))
+model.add(MaxPooling2D((2, 2)))
+model.add(Conv2D(64, (3, 3), activation='relu'))
+model.add(MaxPooling2D((2, 2)))
+model.add(Flatten())
+model.add(Dense(len(unique_labels), activation='softmax'))
+
+# Compile Model
+model.compile(loss='categorical_crossentropy', optimizer=Adam(), metrics=['accuracy'])
+
+# Train Model
+model.fit(X_train, y_train, epochs=10, batch_size=32, validation_data=(X_test, y_test))
+
+# Predictions
+predictions = model.predict(X_test)
+predicted_labels = np.argmax(predictions, axis=1)
+
+# Evaluation
+accuracy = model.evaluate(X_test, y_test)[1]
+print("Accuracy:", accuracy)
+
+# Visualize Spectrograms
+plt.figure(figsize=(10, 5))
+librosa.display.specshow(X_test[0], sr=32000, x_axis='time', y_axis='mel')
+plt.colorbar(format='%+2.0f dB')
+plt.title('Example Spectrogram')
+plt.show()
+
+# Create Submission File (Modify according to the competition's submission format)
+submission_df = pd.DataFrame({'row_id': [i for i in range(len(predictions))],
+                              'predictions': predictions})
+submission_df.to_csv('submission.csv', index=False)

model.joblib

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aa8b5fda6eda56d8b26f1d788e1d0cfb517f4d476341dada66578871a57f798e
+size 32160877

requirements.txt

@@ -0,0 +1,8 @@
+numpy
+pandas
+joblib
+librosa
+wikipedia
+gradio
+plotly
+xgboost

sounds/1000170626.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:20a7f46b20c5be40da57e0f2a58907f75aa99933bc7f8cb3f38e7ab9187c1d28
+size 1834885