update content with the text model from Pierre-Loic repository https://huggingface.co/spaces/pierre-loic/submission

model_audio.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ffe51d1bdc42eab349629b4e178180c9f345509b333eaa05d3e3ff29d71efeb3
+size 6345

tasks/audio.py

@@ -1,3 +1,7 @@
+import librosa
+import joblib
+import numpy as np
+
 from fastapi import APIRouter
 from datetime import datetime
 from datasets import load_dataset
@@ -6,14 +10,14 @@ import random
 import os
 
 from .utils.evaluation import AudioEvaluationRequest
-from .utils.emissions import get_tracker, clean_emissions_data, get_space_info
+from .utils.emissions import tracker, clean_emissions_data, get_space_info
 
 from dotenv import load_dotenv
 load_dotenv()
 
 router = APIRouter()
 
-DESCRIPTION = "Random Baseline"
+DESCRIPTION = "Decision tree"
 ROUTE = "/audio"
 
 
@@ -24,9 +28,7 @@ async def evaluate_audio(request: AudioEvaluationRequest):
     """
     Evaluate audio classification for rainforest sound detection.
     
-    Current Model: Random Baseline
-    - Makes random predictions from the label space (0-1)
-    - Used as a baseline for comparison
+    Current Model: Basic decision tree
     """
     # Get space info
     username, space_url = get_space_info()
@@ -45,19 +47,28 @@ async def evaluate_audio(request: AudioEvaluationRequest):
     test_dataset = train_test["test"]
     
     # Start tracking emissions
-    tracker = get_tracker()
     tracker.start()
     tracker.start_task("inference")
     
     #--------------------------------------------------------------------------------------------
-    # YOUR MODEL INFERENCE CODE HERE
-    # Update the code below to replace the random baseline by your model inference within the inference pass where the energy consumption and emissions are tracked.
+    # MY MODEL
     #--------------------------------------------------------------------------------------------   
-    
-    # Make random predictions (placeholder for actual model inference)
-    true_labels = test_dataset["label"]
-    predictions = [random.randint(0, 1) for _ in range(len(true_labels))]
-    
+    def extract_features(example, sampling_rate):
+        audio_array = example['audio']['array']
+        # mfcc = librosa.feature.mfcc(y=audio_array, sr=sampling_rate, n_mfcc=5)
+        mfcc = librosa.feature.spectral_contrast(y=audio_array)
+        return np.mean(mfcc, axis=1)
+
+    def predict_new_audio(model, dataset, sampling_rate):
+        features_list = [extract_features(example, sampling_rate) for example in dataset]
+        features_array = np.vstack(features_list)
+        predictions = model.predict(features_array)
+        return predictions
+
+    model_filename = "model_audio.pkl"
+    clf = joblib.load(model_filename)
+
+    predictions = predict_new_audio(clf, test_dataset, 12000)
     #--------------------------------------------------------------------------------------------
     # YOUR MODEL INFERENCE STOPS HERE
     #--------------------------------------------------------------------------------------------   
@@ -66,6 +77,7 @@ async def evaluate_audio(request: AudioEvaluationRequest):
     emissions_data = tracker.stop_task()
     
     # Calculate accuracy
+    true_labels = test_dataset["label"]
     accuracy = accuracy_score(true_labels, predictions)
     
     # Prepare results dictionary
@@ -86,4 +98,4 @@ async def evaluate_audio(request: AudioEvaluationRequest):
         }
     }
     
-    return results 
+    return results