Create app.py

app.py

@@ -0,0 +1,173 @@
+import os
+import random
+import warnings
+import logging
+from fastapi import FastAPI, File, UploadFile
+from fastapi.middleware.cors import CORSMiddleware
+from PIL import Image
+import io
+import torch
+import numpy as np
+import cv2
+import mediapipe as mp
+from transformers import Owlv2Processor, Owlv2ForObjectDetection
+from transformers import CLIPProcessor, CLIPModel
+from duckduckgo_search import DDGS
+
+# --- AYARLAR ---
+warnings.filterwarnings("ignore")
+logging.getLogger("transformers").setLevel(logging.ERROR)
+
+app = FastAPI()
+
+# Mobil uygulamadan gelen isteklere izin ver
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],
+    allow_credentials=True,
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+# --- MODELLERİ BAŞLATTA YÜKLE ---
+print("⏳ Modeller Yükleniyor...")
+device = "cpu" # Hugging Face Free Tier CPU kullanır
+
+# OWL-v2
+owl_id = "google/owlv2-base-patch16-ensemble"
+owl_processor = Owlv2Processor.from_pretrained(owl_id)
+owl_model = Owlv2ForObjectDetection.from_pretrained(owl_id).to(device)
+
+# CLIP
+clip_id = "openai/clip-vit-base-patch32"
+clip_processor = CLIPProcessor.from_pretrained(clip_id)
+clip_model = CLIPModel.from_pretrained(clip_id).to(device)
+
+# MediaPipe
+mp_face_mesh = mp.solutions.face_mesh
+face_mesh = mp_face_mesh.FaceMesh(static_image_mode=True, max_num_faces=1, refine_landmarks=True)
+
+print("✅ Sunucu Hazır!")
+
+# --- SÖZLÜKLER VE YEDEK ÜRÜNLER ---
+TR_LABELS = {
+    "acne": "Akne", "pimple": "Sivilce", "dark spot": "Leke",
+    "wrinkles": "Kırışıklık", "oily skin": "Yağlanma",
+    "dry flaky skin": "Kuruluk", "skin redness": "Kızarıklık",
+    "peeling skin": "Soyulma", "rough skin": "Pürüzlü"
+}
+
+FALLBACK_DATABASE = {
+    "Salisilik Asit": [{"title": "La Roche-Posay Effaclar", "link": "https://www.trendyol.com/sr?q=effaclar"}, {"title": "CeraVe SA", "link": "https://www.hepsiburada.com/ara?q=cerave+sa"}],
+    "Çay Ağacı": [{"title": "The Body Shop Tea Tree", "link": "https://www.thebodyshop.com.tr"}, {"title": "Sebamed Clear Face", "link": "https://www.gratis.com"}],
+    "C Vitamini": [{"title": "Garnier C Vitamini", "link": "https://www.trendyol.com"}, {"title": "La Roche-Posay C10", "link": "https://www.hepsiburada.com"}],
+    "Hyaluronik Asit": [{"title": "L'Oreal Hyaluron Uzmanı", "link": "https://www.trendyol.com"}, {"title": "Vichy Mineral 89", "link": "https://www.hepsiburada.com"}],
+    "Centella": [{"title": "Dr. Jart+ Cicapair", "link": "https://www.sephora.com.tr"}, {"title": "Missha Cica", "link": "https://www.missha.com.tr"}]
+}
+
+# --- FONKSİYONLAR ---
+def get_skin_type(image):
+    prompts = ["extremely oily shiny skin", "very dry flaky skin", "normal skin", "combination skin"]
+    labels = ["YAĞLI", "KURU", "NORMAL", "KARMA"]
+    inputs = clip_processor(text=prompts, images=image, return_tensors="pt", padding=True).to(device)
+    with torch.no_grad(): probs = clip_model(**inputs).logits_per_image.softmax(dim=1)
+    return labels[torch.max(probs, 1).indices.item()]
+
+def get_products(ingredient, skin_type):
+    found = []
+    # Canlı Arama
+    try:
+        with DDGS() as ddgs:
+            query = f"site:trendyol.com {ingredient} {skin_type} cilt"
+            results = list(ddgs.text(query, max_results=2))
+            for r in results:
+                found.append({"title": r['title'].split("|")[0], "link": r['href'], "source": "Trendyol"})
+    except: pass
+    
+    # Yedek Depo
+    if len(found) < 2:
+        key_found = None
+        for key in FALLBACK_DATABASE:
+            if key in ingredient: key_found = key
+        if key_found:
+            for item in FALLBACK_DATABASE[key_found]:
+                found.append({"title": item['title'], "link": item['link'], "source": "Öneri"})
+    
+    return found[:4]
+
+def generate_prescription(skin_type, issues):
+    prescriptions = []
+    # Temel
+    if skin_type == "YAĞLI": prescriptions.append({"sorun": "Yağlı Cilt Temizliği", "icerik": "Salisilik Asit"})
+    elif skin_type == "KURU": prescriptions.append({"sorun": "Kuru Cilt Onarımı", "icerik": "Hyaluronik Asit"})
+    
+    # Soruna Özel
+    unique_issues = list(set(issues))
+    for issue in unique_issues:
+        if "acne" in issue or "pimple" in issue:
+            prescriptions.append({"sorun": "Akne Tedavisi", "icerik": "Çay Ağacı"})
+            break
+        elif "dark spot" in issue:
+            prescriptions.append({"sorun": "Leke Giderici", "icerik": "C Vitamini"})
+        elif "redness" in issue:
+            prescriptions.append({"sorun": "Kızarıklık Giderici", "icerik": "Centella"})
+            
+    return prescriptions
+
+# --- API ENDPOINT ---
+@app.get("/")
+def home():
+    return {"status": "Pure Sense API Çalışıyor", "version": "1.0"}
+
+@app.post("/analyze")
+async def analyze_skin(file: UploadFile = File(...)):
+    contents = await file.read()
+    image = Image.open(io.BytesIO(contents)).convert("RGB")
+    
+    # 1. Cilt Tipi
+    skin_type = get_skin_type(image)
+    
+    # 2. Sorun Tespiti
+    text_queries = [["acne", "pimple", "dark spot", "skin redness", "dry flaky skin"]]
+    inputs = owl_processor(text=text_queries, images=image, return_tensors="pt").to(device)
+    with torch.no_grad(): outputs = owl_model(**inputs)
+    
+    target_sizes = torch.Tensor([image.size[::-1]])
+    results = owl_processor.post_process_object_detection(outputs=outputs, target_sizes=target_sizes, threshold=0.02)[0]
+    
+    detections = []
+    issues_found = []
+    
+    for score, label, box in zip(results["scores"], results["labels"], results["boxes"]):
+        lbl_en = text_queries[0][label]
+        conf = round(score.item() * 100, 1)
+        
+        if lbl_en in ["acne", "pimple"] and conf < 10: continue
+        if lbl_en not in ["acne", "pimple"] and conf < 3: continue
+        
+        lbl_tr = TR_LABELS.get(lbl_en, lbl_en)
+        if lbl_tr not in issues_found: issues_found.append(lbl_tr)
+        
+        detections.append({
+            "label": lbl_tr,
+            "confidence": conf,
+            "box": [int(i) for i in box.tolist()]
+        })
+    
+    # 3. Reçete
+    prescriptions = []
+    rx_list = generate_prescription(skin_type, [i.lower() for i in text_queries[0] if TR_LABELS.get(i,i) in issues_found]) # Basit eşleştirme
+    
+    for rx in rx_list:
+        prods = get_products(rx['icerik'], skin_type)
+        prescriptions.append({
+            "title": rx['sorun'],
+            "ingredient": rx['icerik'],
+            "products": prods
+        })
+
+    return {
+        "skin_type": skin_type,
+        "detections": detections,
+        "prescriptions": prescriptions
+    }