Update app.py

app.py

@@ -1,12 +1,12 @@
 # app.py
-# Accent Detection + Cuisine Recommendation (MFCC + HuBERT)
+# Accent Detection + Regional Cuisine Recommendation (MFCC + HuBERT)
 
 import os
 from pathlib import Path
 import pickle
 import numpy as np
 import librosa
-import soundfile as sf  # imported so soundfile is available if needed
+import soundfile as sf  # noqa: F401
 
 from sklearn.preprocessing import StandardScaler, LabelEncoder
 from sklearn.linear_model import LogisticRegression  # noqa: F401
@@ -23,96 +23,209 @@ except Exception:
 DEFAULT_SR = 16000
 CHECKPOINT_DIR = Path("checkpoints")
 
-# ---------- Cuisine mapping ----------
+# ---------------------------------------------------------------------
+# 1. CUISINE MAP (based on your exact dishes)
+# ---------------------------------------------------------------------
 
 CUISINE_MAP = {
-    "hindi": {
-        "region": "North India",
-        "starters_veg": ["Paneer Tikka", "Veg Samosa"],
-        "starters_nonveg": ["Chicken 65", "Mutton Seekh Kebab"],
-        "main_veg": ["Paneer Butter Masala", "Dal Makhani", "Aloo Paratha"],
-        "main_nonveg": ["Butter Chicken", "Rogan Josh", "Mutton Biryani"],
-        "desserts": ["Gulab Jamun", "Kheer"],
-    },
-    "telugu": {
-        "region": "Andhra Pradesh / Telangana",
-        "starters_veg": ["Mirchi Bajji", "Pesarattu"],
-        "starters_nonveg": ["Gongura Chicken Fry", "Fish Fry"],
-        "main_veg": ["Pappu with Rice", "Gutti Vankaya", "Pulihora"],
-        "main_nonveg": ["Andhra Chicken Biryani", "Kodi Pulusu"],
-        "desserts": ["Bobbatlu", "Pootharekulu"],
-    },
-    "tamil": {
-        "region": "Tamil Nadu",
-        "starters_veg": ["Medu Vada", "Sundal"],
-        "starters_nonveg": ["Chicken 65 (TN style)", "Fish Fry"],
-        "main_veg": ["Idli & Sambar", "Masala Dosa", "Rasam Rice"],
-        "main_nonveg": ["Chettinad Chicken", "Fish Curry (Meen Kuzhambu)"],
-        "desserts": ["Payasam", "Kesari"],
+    "andhra_pradesh": {
+        "region": "Andhra Pradesh",
+        "starters_veg": [
+            "Pesarattu (green gram dosa)"
+        ],
+        "starters_nonveg": [
+            "Kodi Vepudu (Andhra chicken fry)",
+            "Royyala Vepudu (prawn fry)"
+        ],
+        "main_veg": [
+            "Pulihora",
+            "Ulava Charu with Rice"
+        ],
+        "main_nonveg": [
+            "Gongura Mutton",
+            "Andhra Chicken Curry / Kodi Kura"
+        ],
+        "desserts": [
+            "Bobbatlu (Puran Poli)",
+            "Pootharekulu",
+            "Paramannam (rice kheer)"
+        ],
     },
+
     "kerala": {
         "region": "Kerala",
-        "starters_veg": ["Banana Chips", "Uzhunnu Vada"],
-        "starters_nonveg": ["Kerala Fish Fry", "Chicken 65 (Kerala Style)"],
-        "main_veg": ["Appam with Veg Stew", "Puttu with Kadala Curry"],
-        "main_nonveg": ["Kerala Fish Curry", "Beef Ularthiyathu", "Chicken Stew"],
-        "desserts": ["Palada Payasam", "Ada Pradhaman"],
+        "starters_veg": [
+            "Banana Chips"
+        ],
+        "starters_nonveg": [
+            "Erachi Fry (Beef Fry)",
+            "Fish Cutlets"
+        ],
+        "main_veg": [
+            "Puttu & Kadala Curry",
+            "Appam with Veg Stew",
+            "Kerala Sadya"
+        ],
+        "main_nonveg": [
+            "Karimeen Pollichathu",
+            "Kerala Fish Curry"
+        ],
+        "desserts": [
+            "Palada Payasam",
+            "Ada Pradhaman",
+            "Elaneer Payasam"
+        ],
     },
-    "kannada": {
-        "region": "Karnataka",
-        "starters_veg": ["Maddur Vada", "Bonda"],
-        "starters_nonveg": ["Kheema Cutlet"],
-        "main_veg": ["Bisi Bele Bath", "Ragi Mudde", "Akki Roti"],
-        "main_nonveg": ["Mangalorean Chicken Curry", "Neer Dosa with Fish Curry"],
-        "desserts": ["Mysore Pak", "Kesari"],
+
+    "gujarat": {
+        "region": "Gujarat",
+        "starters_veg": [
+            "Dhokla",
+            "Khandvi",
+            "Sev Khamani"
+        ],
+        "starters_nonveg": [
+            "Local non‑veg starters (rare in traditional cuisine)"
+        ],
+        "main_veg": [
+            "Undhiyu",
+            "Thepla",
+            "Dal Dhokli",
+            "Sev Tameta Nu Shaak"
+        ],
+        "main_nonveg": [
+            "Local non‑veg mains (if available)"
+        ],
+        "desserts": [
+            "Basundi",
+            "Shrikhand",
+            "Mohanthal"
+        ],
+    },
+
+    "jharkhand": {
+        "region": "Jharkhand",
+        "starters_veg": [
+            "Dhuska",
+            "Chilka Roti",
+            "Rugra Fry (mushroom fry)"
+        ],
+        "starters_nonveg": [
+            "Local non‑veg starters"
+        ],
+        "main_veg": [
+            "Bamboo Shoot Curry",
+            "Kadho (local dal curry)",
+            "Litti‑Chokha"
+        ],
+        "main_nonveg": [
+            "Local non‑veg curries",
+            "Handia (served with meals)"
+        ],
+        "desserts": [
+            "Thekua",
+            "Tilkut",
+            "Malpua (Jharkhand style)"
+        ],
     },
-    "bengali": {
-        "region": "West Bengal",
-        "starters_veg": ["Vegetable Chop"],
-        "starters_nonveg": ["Fish Kabiraji", "Chicken Pakora"],
-        "main_veg": ["Shukto", "Cholar Dal with Luchi"],
-        "main_nonveg": ["Machher Jhol", "Ilish Bhapa"],
-        "desserts": ["Mishti Doi", "Rasgulla"],
+
+    "tamil_nadu": {
+        "region": "Tamil Nadu",
+        "starters_veg": [
+            "Medu Vada",
+            "Masala Vadai"
+        ],
+        "starters_nonveg": [
+            "Chicken 65"
+        ],
+        "main_veg": [
+            "Sambar Rice",
+            "Pongal"
+        ],
+        "main_nonveg": [
+            "Chettinad Chicken Curry",
+            "Kothu Parotta"
+        ],
+        "desserts": [
+            "Payasam",
+            "Kesari",
+            "Jigarthanda (Madurai)"
+        ],
     },
+
+    "karnataka": {
+        "region": "Karnataka",
+        "starters_veg": [
+            "Maddur Vada",
+            "Goli Baje"
+        ],
+        "starters_nonveg": [
+            "Mangalore Chicken Ghee Roast"
+        ],
+        "main_veg": [
+            "Bisi Bele Bath",
+            "Neer Dosa",
+            "Ragi Mudde with Sambar"
+        ],
+        "main_nonveg": [
+            "Coorg Pandi Curry"
+        ],
+        "desserts": [
+            "Mysore Pak",
+            "Kesari Bath",
+            "Obbattu / Holige"
+        ],
+    },
+
+    # Fallback if something isn't mapped yet
     "default": {
         "region": "Unknown / Other",
         "starters_veg": ["Local vegetarian starters"],
-        "starters_nonveg": ["Local non-veg starters"],
+        "starters_nonveg": ["Local non‑veg starters"],
         "main_veg": ["Local vegetarian mains"],
-        "main_nonveg": ["Local non-veg mains"],
+        "main_nonveg": ["Local non‑veg mains"],
         "desserts": ["Local desserts"],
     },
 }
 
-# Map various predicted labels / codes to cuisine keys
+# ---------------------------------------------------------------------
+# 2. LABEL ALIAS MAP (so HuBERT/MFCC labels map to these keys)
+# ---------------------------------------------------------------------
+
 LABEL_ALIAS = {
-    # Hindi
-    "hi": "hindi",
-    "hindi": "hindi",
-    "hindi_india": "hindi",
-    # Telugu
-    "te": "telugu",
-    "telugu": "telugu",
-    "telugu_india": "telugu",
-    # Tamil
-    "ta": "tamil",
-    "tamil": "tamil",
-    "tamil_india": "tamil",
-    # Bengali
-    "bn": "bengali",
-    "bengali": "bengali",
-    "bengali_india": "bengali",
-    # Kannada
-    "kn": "kannada",
-    "kannada": "kannada",
-    "kannada_india": "kannada",
-    # Malayalam / Kerala
+    # Andhra / Telugu-ish labels
+    "andhra": "andhra_pradesh",
+    "andhra_pradesh": "andhra_pradesh",
+    "ap": "andhra_pradesh",
+    "telugu": "andhra_pradesh",
+    "telugu_andhra": "andhra_pradesh",
+    "telugu_india": "andhra_pradesh",
+
+    # Kerala / Malayalam
+    "kerala": "kerala",
     "ml": "kerala",
     "malayalam": "kerala",
-    "kerala": "kerala",
     "kerala_malayalam": "kerala",
-}
 
+    # Gujarat
+    "gujarat": "gujarat",
+    "gj": "gujarat",
+    "gujarati": "gujarat",
+
+    # Jharkhand
+    "jharkhand": "jharkhand",
+    "jh": "jharkhand",
+
+    # Tamil Nadu
+    "tamil_nadu": "tamil_nadu",
+    "tamil": "tamil_nadu",
+    "tn": "tamil_nadu",
+
+    # Karnataka
+    "karnataka": "karnataka",
+    "ka": "karnataka",
+    "kannada": "karnataka",
+}
 
 def format_cuisine_output(pred_label: str) -> str:
     # Normalise raw predicted label and map through alias
@@ -131,8 +244,9 @@ def format_cuisine_output(pred_label: str) -> str:
     lines.append("Desserts: " + ", ".join(info["desserts"]))
     return "\n".join(lines)
 
-
-# ---------- Audio & feature helpers ----------
+# ---------------------------------------------------------------------
+# 3. Audio & Feature helpers
+# ---------------------------------------------------------------------
 
 def normalize_audio_path(audio):
     if audio is None:
@@ -160,7 +274,7 @@ _hf_model = None
 
 def get_hubert_layer_embedding(path: str, layer_idx: int = 11) -> np.ndarray:
     """
-    Uses facebook/hubert-large-ll60k (hidden size 1024, matching your scaler_hubert.pkl).
+    Uses facebook/hubert-large-ll60k (hidden size 1024, matching a 1024‑dim scaler).
     Returns a 1024‑dim vector: mean‑pooled over time from the chosen layer.
     """
     global _hf_feat, _hf_model
@@ -180,7 +294,7 @@ def get_hubert_layer_embedding(path: str, layer_idx: int = 11) -> np.ndarray:
     with torch.no_grad():
         out = _hf_model(**inputs)
 
-    hidden_states = out.hidden_states  # list[Tensor(batch,time,dim)]
+    hidden_states = out.hidden_states  # list[Tensor(batch, time, dim)]
     hs = hidden_states[layer_idx]      # (batch, time, dim)
 
     if hs.ndim == 3:
@@ -189,14 +303,14 @@ def get_hubert_layer_embedding(path: str, layer_idx: int = 11) -> np.ndarray:
         vec = hs.mean(dim=0).cpu().numpy()
     return vec
 
-
-# ---------- Load models from checkpoints ----------
+# ---------------------------------------------------------------------
+# 4. Load models from checkpoints
+# ---------------------------------------------------------------------
 
 app_state = {"models": {}}
 
-
 def load_models():
-    # MFCC
+    # MFCC model
     try:
         with open(CHECKPOINT_DIR / "clf_mfcc.pkl", "rb") as f:
             clf_mfcc = pickle.load(f)
@@ -213,7 +327,7 @@ def load_models():
     except Exception as e:
         print("Could not load MFCC model:", e)
 
-    # HuBERT
+    # HuBERT model
     try:
         with open(CHECKPOINT_DIR / "clf_hubert.pkl", "rb") as f:
             clf_h = pickle.load(f)
@@ -236,11 +350,11 @@ def load_models():
     except Exception as e:
         print("Could not load HuBERT model:", e)
 
-
 load_models()
 
-
-# ---------- Prediction logic ----------
+# ---------------------------------------------------------------------
+# 5. Prediction logic
+# ---------------------------------------------------------------------
 
 def predict_accent_and_cuisine(audio_file, feature_choice, hubert_layer_idx, use_trained):
     audio_path = normalize_audio_path(audio_file)
@@ -269,7 +383,7 @@ def predict_accent_and_cuisine(audio_file, feature_choice, hubert_layer_idx, use
         if feat == "mfcc":
             x = extract_mfcc_pooled(audio_path)
         else:
-            # use trained HuBERT layer if saved, otherwise slider value
+            # Use stored trained layer if available; otherwise the slider value
             trained_layer = model_info.get("layer", hubert_layer_idx)
             x = get_hubert_layer_embedding(audio_path, layer_idx=trained_layer)
     except Exception as e:
@@ -286,8 +400,9 @@ def predict_accent_and_cuisine(audio_file, feature_choice, hubert_layer_idx, use
     cuisine_text = format_cuisine_output(pred_label)
     return pred_label, cuisine_text
 
-
-# ---------- Gradio UI ----------
+# ---------------------------------------------------------------------
+# 6. Gradio UI
+# ---------------------------------------------------------------------
 
 with gr.Blocks() as demo:
     gr.Markdown("# Accent Detection & Cuisine Recommendation")