Create 안녕

안녕

@@ -0,0 +1,75 @@
+pip install transformers torch pillow requests
+from transformers import BlipProcessor, BlipForQuestionAnswering
+from PIL import Image
+import torch
+import requests
+from io import BytesIO
+
+class VQASystem:
+    def __init__(self, model_name="Salesforce/blip-vqa-base"):
+        """VQA 모델 초기화"""
+        print(f"🔧 VQA 모델 로드 중: {model_name}")
+        self.processor = BlipProcessor.from_pretrained(model_name)
+        self.model = BlipForQuestionAnswering.from_pretrained(model_name)
+        self.device = "cuda" if torch.cuda.is_available() else "cpu"
+        self.model.to(self.device)
+        print("✅ 모델 로드 완료")
+
+    def load_image(self, image_source):
+        """이미지 로드"""
+        if image_source.startswith('http'):
+            response = requests.get(image_source)
+            image = Image.open(BytesIO(response.content)).convert('RGB')
+        else:
+            image = Image.open(image_source).convert('RGB')
+        return image
+
+    def generate_answer(self, image_path, question):
+        """질문에 대한 답변 생성"""
+        try:
+            raw_image = self.load_image(image_path)
+            
+            # 모델 입력 생성
+            inputs = self.processor(raw_image, question, return_tensors="pt").to(self.device)
+            
+            # 답변 생성 (max_new_tokens 조절 가능)
+            with torch.no_grad():
+                out = self.model.generate(**inputs, max_new_tokens=50)
+                
+            answer = self.processor.decode(out[0], skip_special_tokens=True)
+            return answer
+        except Exception as e:
+            return f"Error: {str(e)}"
+
+    def batch_qa(self, image_path, questions):
+        """여러 질문 일괄 처리"""
+        print(f"🖼️ 이미지 분석 중: {image_path}")
+        results = {}
+        for q in questions:
+            ans = self.generate_answer(image_path, q)
+            results[q] = ans
+            print(f"Q: {q}\nA: {ans}\n")
+        return results
+
+def main():
+    print("="*60)
+    print("Project 2: Visual Question Answering System")
+    print("="*60)
+    
+    vqa = VQASystem()
+    
+    # 테스트용 이미지
+    test_image = "https://raw.githubusercontent.com/pytorch/hub/master/images/dog.jpg"
+    
+    questions = [
+        "What animal is in the picture?",
+        "What is the dog doing?",
+        "What color is the dog?",
+        "Is there a cat in the image?"
+    ]
+    
+    vqa.batch_qa(test_image, questions)
+
+if __name__ == "__main__":
+    main()
+