Update app.py

app.py

@@ -1,13 +1,11 @@
 import os
 import io
-import numpy as np
 from PIL import Image
 from flask import Flask, request, jsonify, send_file
 from flask_cors import CORS 
 import re 
 import torch 
-from transformers import AutoModelForImageClassification, AutoImageProcessor
-from typing import Dict, Any
+from transformers import pipeline
 
 # =================================================================
 # 1. SETUP FLASK SERVER & CORS
@@ -16,95 +14,46 @@ app = Flask(__name__)
 CORS(app) 
 
 # =================================================================
-# 2. SETUP MODEL AI & PREPROCESSING
+# 2. SETUP MODEL AI (Kembali ke Pipeline Sederhana)
 # =================================================================
 MODEL_NAME = "prithivMLmods/BrainTumor-Classification-Mini"
-model = None 
-processor = None
-device = "cpu" # Default ke CPU untuk stabilitas Hugging Face Spaces
+classifier = None 
 
 def load_model():
-    """Memuat model dan image processor secara eksplisit untuk menjamin preprocessing yang benar."""
-    global model, processor, device
+    """Memuat model AI menggunakan pipeline standar."""
+    global classifier
     try:
-        print("⏳ Sedang memuat Model dan Processor AI ({})...".format(MODEL_NAME))
+        print("⏳ Sedang memuat model AI ({})...".format(MODEL_NAME))
         device = "cuda" if torch.cuda.is_available() else "cpu"
-
-        # Memuat Processor (WAJIB untuk Preprocessing yang Benar)
-        processor = AutoImageProcessor.from_pretrained(MODEL_NAME)
-        
-        # Memuat Model
-        model = AutoModelForImageClassification.from_pretrained(MODEL_NAME)
-        model.to(device)
-        model.eval() # Atur model ke mode evaluasi
-        
-        print(f"✅ Model dan Processor berhasil dimuat ke device: {device}")
+        classifier = pipeline(
+            "image-classification",
+            model=MODEL_NAME,
+            device=device 
+        )
+        print("✅ Model {} berhasil dimuat ke device: {}".format(MODEL_NAME, device))
     except Exception as e:
-        print("❌ Gagal memuat model atau processor: {}".format(e))
-        model = None 
-        processor = None
+        print("❌ Gagal memuat model: {}".format(e))
+        classifier = None 
 
-# Panggil fungsi saat aplikasi dimuat oleh Gunicorn
 load_model()
 
-# Pemetaan label (diambil dari model's config, jika ada)
-LABEL_MAPPING_RAW = {
+# Pemetaan label
+LABEL_MAPPING = {
     'glioma': {'status': 'Tumor Terdeteksi', 'jenis': 'Glioma'},
     'meningioma': {'status': 'Tumor Terdeteksi', 'jenis': 'Meningioma'},
     'notumor': {'status': 'Tidak Ada Tumor', 'jenis': 'Tidak Ada'},
     'pituitary': {'status': 'Tumor Terdeteksi', 'jenis': 'Pituitary Tumor'}
 }
 
-# =================================================================
-# 3. FUNGSI PREDIKSI MANUAL (untuk Akurasi Tertinggi)
-# =================================================================
-
-def predict_manual(image: Image.Image) -> Dict[str, Any]:
-    """Mengambil gambar PIL, mengolahnya dengan processor, dan memprediksi secara manual."""
-    if model is None or processor is None:
-        raise Exception("Model atau Processor belum siap.")
-        
-    # 1. Preprocessing Gambar menggunakan processor model
-    inputs = processor(image, return_tensors="pt")
-    
-    # 2. Memindahkan input ke device yang benar (CPU/GPU)
-    inputs = {k: v.to(device) for k, v in inputs.items()}
-
-    # 3. Inference (prediksi)
-    with torch.no_grad():
-        outputs = model(**inputs)
-        logits = outputs.logits
-    
-    # 4. Ambil Probabilitas dan Index Tertinggi
-    probabilities = torch.nn.functional.softmax(logits, dim=-1)
-    top_prob, top_index = torch.topk(probabilities, 1)
-
-    # 5. Dapatkan Label
-    # Mengambil label dari ID ke string (jika ada di model config)
-    predicted_id = top_index.item()
-    if model.config.id2label:
-        raw_label = model.config.id2label[predicted_id]
-    else:
-        # Fallback jika id2label tidak ada (seperti pada prithivMLmods)
-        # Kita harus berasumsi urutan index adalah: 0=glioma, 1=meningioma, 2=notumor, 3=pituitary
-        index_to_label = ['glioma', 'meningioma', 'notumor', 'pituitary']
-        raw_label = index_to_label[predicted_id] if predicted_id < len(index_to_label) else 'unknown'
-
-    # 6. Formatting Output
-    clean_key = raw_label.lower().replace(' ', '')
-    result_data = LABEL_MAPPING_RAW.get(clean_key, {
-        'status': 'Hasil Tidak Dikenal', 
-        'jenis': 'N/A'
-    })
-    
-    return {
-        "prediction_status": result_data['status'],
-        "tumor_type": result_data['jenis'],
-        "confidence": round(top_prob.item() * 100, 2)
-    }
+def clean_label(raw_label):
+    """Membersihkan label mentah dari model."""
+    match = re.search(r'(glioma|meningioma|notumor|pituitary)', raw_label, re.IGNORECASE)
+    if match:
+        return match.group(0).lower()
+    return raw_label.lower()
 
 # =================================================================
-# 4. ENDPOINT WEB SERVER (Host HTML)
+# 3. ENDPOINT WEB SERVER (Host HTML)
 # =================================================================
 
 @app.route('/', methods=['GET'])
@@ -116,32 +65,47 @@ def serve_html():
         return f"<h1>Error: File index.html tidak ditemukan di server.</h1><p>Pastikan file index.html sudah di-COPY ke Docker container.</p><p>{e}</p>", 500
 
 # =================================================================
-# 5. ENDPOINT API (Predict)
+# 4. ENDPOINT API (Predict)
 # =================================================================
 
 @app.route('/predict', methods=['POST'])
 def predict():
-    """Endpoint utama untuk prediksi, menggunakan fungsi predict_manual yang akurat."""
-    
+    """Endpoint utama untuk prediksi."""
+    if classifier is None:
+        return jsonify({"error": "Model AI belum dimuat atau gagal dimuat."}), 500
+
     if 'file' not in request.files:
         return jsonify({"error": "no file uploaded"}), 400
     
     file = request.files['file']
     
     try:
-        # 1. Baca file gambar dan konversi ke PIL Image
         image_bytes = file.read()
         image = Image.open(io.BytesIO(image_bytes)).convert("RGB")
         
     except Exception as e:
-        print(f"Error saat memproses gambar: {e}")
         return jsonify({"error": f"Gagal memproses gambar: {e}"}), 400
 
-    # Lakukan Prediksi Manual
+    # Lakukan Prediksi
     try:
-        result = predict_manual(image)
-        return jsonify(result)
+        # Gunakan pipeline
+        results = classifier(images=image, top_k=1) 
+        
+        result = results[0]
+        raw_label = result['label']
+        confidence = result['score'] * 100
+
+        clean_key = clean_label(raw_label)
+        result_data = LABEL_MAPPING.get(clean_key, {
+            'status': 'Hasil Tidak Dikenal', 
+            'jenis': 'N/A'
+        })
+        
+        return jsonify({
+            "prediction_status": result_data['status'],
+            "tumor_type": result_data['jenis'],
+            "confidence": round(confidence, 2)
+        })
 
     except Exception as e:
-        print(f"Error saat prediksi: {e}")
-        return jsonify({"error": f"Terjadi kesalahan saat menjalankan prediksi model: {e}"}), 500
+        return jsonify({"error": "Terjadi kesalahan saat menjalankan prediksi model."}), 500