first working app.py

app.py

@@ -0,0 +1,139 @@
+import os
+import zipfile
+import requests
+import numpy as np
+import pandas as pd
+import librosa
+import torch
+import gradio as gr
+
+from transformers import AutoTokenizer, AutoModel
+from sklearn.ensemble import RandomForestClassifier
+from sklearn.preprocessing import StandardScaler
+
+# =========================
+# 1. DOWNLOAD DATASET
+# =========================
+url = "https://huggingface.co/datasets/ananyakarn/DAIC_WOZ_Data/resolve/main/DAIC_WOZ_Data.zip"
+zip_path = "data.zip"
+extract_path = "DAIC_WOZ"
+
+if not os.path.exists(extract_path):
+    print("Downloading dataset...")
+    r = requests.get(url, stream=True)
+    with open(zip_path, "wb") as f:
+        for chunk in r.iter_content(8192):
+            f.write(chunk)
+
+    print("Extracting dataset...")
+    with zipfile.ZipFile(zip_path, "r") as zip_ref:
+        zip_ref.extractall(extract_path)
+
+# =========================
+# 2. LOAD BERT
+# =========================
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
+tokenizer = AutoTokenizer.from_pretrained("bert-base-multilingual-cased")
+bert = AutoModel.from_pretrained("bert-base-multilingual-cased").to(device)
+bert.eval()
+
+# =========================
+# FEATURE FUNCTIONS
+# =========================
+def get_text_embedding(text):
+    inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True, max_length=128).to(device)
+    with torch.no_grad():
+        outputs = bert(**inputs)
+    return outputs.last_hidden_state.mean(dim=1).squeeze().cpu().numpy()
+
+def get_audio_features(folder):
+    try:
+        file = [f for f in os.listdir(folder) if f.endswith("_AUDIO.wav")][0]
+        y, sr = librosa.load(os.path.join(folder, file), sr=16000)
+        mfcc = librosa.feature.mfcc(y=y, sr=sr, n_mfcc=20)
+        return np.mean(mfcc.T, axis=0)
+    except:
+        return np.zeros(20)
+
+def load_text(folder):
+    try:
+        file = [f for f in os.listdir(folder) if "TRANSCRIPT" in f][0]
+        df = pd.read_csv(os.path.join(folder, file))
+        return " ".join(df.iloc[:, -1].astype(str).tolist())
+    except:
+        return ""
+
+def get_label(folder):
+    return 1 if "_P" in folder else 0
+
+# =========================
+# 3. TRAIN MODEL (LIGHT)
+# =========================
+print("Training model...")
+
+X, y = [], []
+folders = os.listdir(extract_path)[:20]  # 🔥 only 20 samples
+
+for folder in folders:
+    path = os.path.join(extract_path, folder)
+
+    if not os.path.isdir(path):
+        continue
+
+    text = load_text(path)
+    text_feat = get_text_embedding(text)
+    audio_feat = get_audio_features(path)
+
+    combined = np.concatenate([text_feat, audio_feat])
+
+    X.append(combined)
+    y.append(get_label(folder))
+
+X = np.array(X)
+y = np.array(y)
+
+scaler = StandardScaler()
+X = scaler.fit_transform(X)
+
+model = RandomForestClassifier(n_estimators=50)
+model.fit(X, y)
+
+print("Model ready!")
+
+# =========================
+# 4. PREDICTION FUNCTION
+# =========================
+def predict(text, audio):
+
+    text_feat = get_text_embedding(text)
+
+    if audio:
+        y_audio, sr = librosa.load(audio, sr=16000)
+        mfcc = librosa.feature.mfcc(y=y_audio, sr=sr, n_mfcc=20)
+        audio_feat = np.mean(mfcc.T, axis=0)
+    else:
+        audio_feat = np.zeros(20)
+
+    x = np.concatenate([text_feat, audio_feat])
+    x = scaler.transform([x])
+
+    pred = model.predict(x)[0]
+
+    return "Depression Detected ⚠️" if pred == 1 else "No Depression ✅"
+
+# =========================
+# 5. UI
+# =========================
+app = gr.Interface(
+    fn=predict,
+    inputs=[
+        gr.Textbox(label="Enter Text"),
+        gr.Audio(type="filepath", label="Upload Audio (optional)")
+    ],
+    outputs="text",
+    title="Multimodal Depression Detection",
+    description="Training + inference demo using text and audio"
+)
+
+app.launch()