adjust metadata logic

main.py

@@ -1,6 +1,5 @@
 from fastapi import FastAPI, File, UploadFile, Form
 from fastapi.middleware.cors import CORSMiddleware
-from typing import Optional
 import numpy as np
 import tensorflow as tf
 from PIL import Image
@@ -21,7 +20,7 @@ def custom_max_pool(x):
 # --- CONFIG ---
 IMG_SIZE = 224
 NUM_CLASSES = 23
-USE_METADATA_ADJUSTMENT = True  # Toggle this to False if you want raw model output
+USE_METADATA_ADJUSTMENT = True  # Toggle on/off
 
 MODEL_PATHS = {
     "EfficientNetB3": "saved_models/EfficientNetB3_recovered.keras",
@@ -62,31 +61,22 @@ app.add_middleware(
     allow_headers=["*"],
 )
 
-# --- HEALTH CHECK ENDPOINT ---
 @app.get("/")
 def root():
     return {"status": "🩺 App is running."}
 
-# --- LOAD MODELS AT STARTUP ---
 models = {}
 
 @app.on_event("startup")
 def load_models():
     global models
-    print("Loading models...")
     os.makedirs("saved_models", exist_ok=True)
-
     for name, path in MODEL_PATHS.items():
         if not os.path.exists(path):
-            print(f"Downloading {name} from Google Drive...")
             gdown.download(MODEL_URLS[name], path, quiet=False)
-
-        print(f"Loading {name}...")
         models[name] = tf.keras.models.load_model(path)
 
-    print("✅ All models loaded.")
-
-# --- IMAGE UTILS ---
+# --- UTILS ---
 def read_imagefile(file) -> Image.Image:
     image = Image.open(io.BytesIO(file))
     return image.convert("RGB")
@@ -112,80 +102,31 @@ def adjust_with_metadata(predictions, metadata):
     race = metadata.get("race", "").lower()
 
     rules = [
-        # 0. Acne and Rosacea
-        {"keyword": "acne", "condition": age < 25 or "oily" in skin_type, "factor": 1.3},
-        {"keyword": "rosacea", "condition": "redness" in condition or "cheek" in condition, "factor": 1.3},
-
-        # 1. AK/Basal Cell/Malignant Lesions
-        {"keyword": "keratosis", "condition": age > 60 or "lesion" in condition, "factor": 1.3},
-        {"keyword": "carcinoma", "condition": "bleeding" in condition or "non-healing" in condition, "factor": 1.3},
-
-        # 2. Atopic Dermatitis
-        {"keyword": "atopic", "condition": age < 10 or "itchy" in condition, "factor": 1.3},
-
-        # 3. Bullous Disease
-        {"keyword": "bullous", "condition": "blisters" in condition or age > 60, "factor": 1.3},
-
-        # 4. Cellulitis/Impetigo/Bacterial
-        {"keyword": "impetigo", "condition": "crust" in condition or "yellow" in condition, "factor": 1.3},
-        {"keyword": "cellulitis", "condition": "swelling" in condition or "fever" in condition, "factor": 1.3},
-
-        # 5. Eczema
-        {"keyword": "eczema", "condition": "dry" in skin_type or "itchy" in condition, "factor": 1.3},
-
-        # 6. Exanthems/Drug Eruption
-        {"keyword": "drug eruption", "condition": "medication" in condition or "rash" in condition, "factor": 1.3},
-
-        # 7. Hair Loss
-        {"keyword": "hair loss", "condition": gender == "male" and age > 30, "factor": 1.3},
-
-        # 8. Herpes/HPV/STDs
-        {"keyword": "herpes", "condition": "painful" in condition or "genital" in condition, "factor": 1.3},
-
-        # 9. Pigmentation Disorders
-        {"keyword": "pigmentation", "condition": skin_type in ["dark", "brown"], "factor": 1.3},
-
-        # 10. Lupus/Connective Tissue
-        {"keyword": "lupus", "condition": "malar" in condition or gender == "female", "factor": 1.3},
-        {"keyword": "connective", "condition": "joint" in condition or "fatigue" in condition, "factor": 1.3},
-
-        # 11. Melanoma/Skin Cancer/Nevi
+        {"keyword": "acne", "condition": age > 40, "factor": 0.6},
+        {"keyword": "acne", "condition": age < 25, "factor": 1.2},
+        {"keyword": "eczema", "condition": skin_type == "dry", "factor": 1.2},
+        {"keyword": "eczema", "condition": skin_type == "oily", "factor": 0.8},
+        {"keyword": "warts", "condition": age < 12, "factor": 1.3},
+        {"keyword": "fungal", "condition": "itchy" in condition, "factor": 1.2},
+        {"keyword": "hair loss", "condition": gender == "male" and age > 40, "factor": 1.3},
+        {"keyword": "lupus", "condition": gender == "female", "factor": 1.2},
         {"keyword": "melanoma", "condition": skin_type == "light" and age > 50, "factor": 1.3},
-
-        # 12. Nail Fungus
-        {"keyword": "nail fungus", "condition": "thickened" in condition or age > 60, "factor": 1.3},
-
-        # 13. Poison Ivy / Contact Dermatitis
-        {"keyword": "poison ivy", "condition": "itchy" in condition or "blister" in condition, "factor": 1.3},
-        {"keyword": "contact dermatitis", "condition": "burning" in condition or "irritant" in condition, "factor": 1.3},
-
-        # 14. Psoriasis / Lichen Planus
-        {"keyword": "psoriasis", "condition": "scaly" in condition or "elbow" in condition, "factor": 1.3},
-
-        # 15. Scabies/Lyme/Bites
-        {"keyword": "scabies", "condition": "night" in condition or "burrow" in condition, "factor": 1.3},
-
-        # 16. Seborrheic Keratoses/Benign Tumors
-        {"keyword": "seborrheic", "condition": age > 60 or "waxy" in condition, "factor": 1.3},
-
-        # 17. Systemic Disease
-        {"keyword": "systemic", "condition": "fatigue" in condition or "multiple" in condition, "factor": 1.3},
-
-        # 18. Fungal (Tinea, Ringworm, etc.)
-        {"keyword": "fungal", "condition": "itchy" in condition or "ring" in condition, "factor": 1.3},
-
-        # 19. Urticaria (Hives)
-        {"keyword": "urticaria", "condition": "hives" in condition or "allergy" in condition, "factor": 1.3},
-
-        # 20. Vascular Tumors
-        {"keyword": "vascular tumor", "condition": age < 5 or "birthmark" in condition, "factor": 1.3},
-
-        # 21. Vasculitis
-        {"keyword": "vasculitis", "condition": "purpura" in condition or "painful" in condition, "factor": 1.3},
-
-        # 22. Warts/Molluscum/Viral
-        {"keyword": "warts", "condition": "raised" in condition or "cauliflower" in condition, "factor": 1.3},
-        {"keyword": "molluscum", "condition": "umbilicated" in condition or age < 12, "factor": 1.3},
+        {"keyword": "nail fungus", "condition": age > 60, "factor": 1.3},
+        {"keyword": "psoriasis", "condition": "flaky" in condition or "dry" in skin_type, "factor": 1.2},
+        {"keyword": "systemic", "condition": "fatigue" in condition or "pain" in condition, "factor": 1.2},
+        {"keyword": "urticaria", "condition": "allergy" in condition or "hives" in condition, "factor": 1.3},
+        {"keyword": "contact dermatitis", "condition": "red" in condition or "burning" in condition, "factor": 1.2},
+        {"keyword": "seborrheic", "condition": age > 60, "factor": 1.2},
+        {"keyword": "bullous", "condition": age > 60, "factor": 1.3},
+        {"keyword": "vasculitis", "condition": "joint" in condition or "swelling" in condition, "factor": 1.2},
+        {"keyword": "atopic dermatitis", "condition": age < 10, "factor": 1.2},
+        {"keyword": "pigmentation", "condition": skin_type == "dark", "factor": 1.3},
+        {"keyword": "hpv", "condition": age > 18, "factor": 1.2},
+        {"keyword": "vascular tumors", "condition": age < 5, "factor": 1.3},
+        {"keyword": "poison ivy", "condition": "rash" in condition or "camping" in condition, "factor": 1.3},
+        {"keyword": "bacterial", "condition": "pus" in condition or "fever" in condition, "factor": 1.3},
+        {"keyword": "drug eruption", "condition": "medication" in condition or "rash" in condition, "factor": 1.3},
+        {"keyword": "connective tissue", "condition": "joint" in condition or "fatigue" in condition, "factor": 1.3},
     ]
 
     adjusted = []
@@ -199,7 +140,6 @@ def adjust_with_metadata(predictions, metadata):
 
     return sorted(adjusted, key=lambda x: x["confidence"], reverse=True)[:3]
 
-
 # --- PREDICT ENDPOINT ---
 @app.post("/predict/")
 async def predict(
@@ -241,7 +181,10 @@ async def predict(
     ]
 
     top3_indices = prediction.argsort()[-3:][::-1]
-    top_preds = [{"label": class_labels[i], "confidence": float(prediction[i])} for i in top3_indices]
+    top_preds = [
+        {"label": class_labels[i], "confidence": float(prediction[i])}
+        for i in top3_indices
+    ]
 
     metadata = {
         "age": age,
@@ -256,6 +199,5 @@ async def predict(
 
     return {
         "prediction": final_preds,
-        "metadata": metadata,
-        "metadata_adjustment_applied": USE_METADATA_ADJUSTMENT
+        "metadata": metadata
     }