Update app.py

app.py

@@ -1,66 +1,37 @@
-import gradio as gr
-from transformers import pipeline
 from PIL import Image
-
-# ① 파이프라인 초기화
-captioner = pipeline("image-to-text", model="Salesforce/blip-image-captioning-base")
-scene_classifier = pipeline(
-    "zero-shot-image-classification",
-    model="openai/clip-vit-base-patch32"
-)
-
-SCENE_LABELS = ["outdoor", "indoor", "beach", "office", "street",
-                "restaurant", "park", "sports", "kitchen", "mountain"]
-TEMPLATES = {
-    "outdoor":    "In this picture, {caption}. It looks like a pleasant outdoor setting, and the subject seems relaxed.",
-    "indoor":     "In this picture, {caption}. It appears to be indoors, perhaps at home or in an office environment.",
-    "beach":      "In this picture, {caption}. It seems to be on a beach, and the atmosphere looks warm and sunny.",
-    "office":     "In this picture, {caption}. It looks like an office scene, with people engaged in work or discussion.",
-    "street":     "In this picture, {caption}. The scene appears to be on a busy street, with vehicles and pedestrians.",
-    "restaurant": "In this picture, {caption}. It looks like a restaurant setting, where people are dining together.",
-    "park":       "In this picture, {caption}. The location seems to be a park, with trees and open space.",
-    "sports":     "In this picture, {caption}. It appears to be a sports activity, showing movement and action.",
-    "kitchen":    "In this picture, {caption}. It seems to be in a kitchen, with cooking utensils visible.",
-    "mountain":   "In this picture, {caption}. The background looks like mountains, suggesting a hiking scene."
-}
+import traceback
 
 def generate_caption(filepaths, choice_index):
-    idx = int(choice_index)
-    img_path = filepaths[idx]
-
-    # PIL로 이미지 로드
-    img = Image.open(img_path).convert("RGB")
-
-    # 1) 원본 캡션
-    out = captioner(img)
-    first = out[0] if isinstance(out, list) else out
-    raw = first.get("generated_text") or first.get("text") or str(first)
-    raw = raw.strip()
-
-    # 2) 장면 분류
-    cls = scene_classifier(img, candidate_labels=SCENE_LABELS)
-    scene = cls["labels"][0]
-
-    # 3) 템플릿 매핑
-    template = TEMPLATES.get(scene, "In this picture, {caption}.")
-    return template.format(caption=raw)
-
-with gr.Blocks() as demo:
-    gr.Markdown("## 📸 TOEIC Part 1: 상황별 사진 묘사")
-    # Multiple files 업로드, filepath로 받기
-    img_inputs = gr.Files(
-        file_count="multiple",
-        type="filepath",
-        label="Upload up to 4 images"
-    )
-    choice = gr.Dropdown(
-        choices=[str(i) for i in range(4)],
-        value="0",
-        label="Which image to describe? (0–3)"
-    )
-    btn = gr.Button("Describe")
-    output = gr.Textbox(label="TOEIC Part 1 Response", lines=4)
-    btn.click(fn=generate_caption, inputs=[img_inputs, choice], outputs=output)
-
-if __name__ == "__main__":
-    demo.launch()
+    try:
+        idx = int(choice_index)
+        img_path = filepaths[idx]
+        print(f"[DEBUG] Selected image path: {img_path}")
+
+        # 1) 이미지 로드
+        img = Image.open(img_path).convert("RGB")
+        print("[DEBUG] Image loaded")
+
+        # 2) 캡션 생성
+        out = captioner(img)
+        print(f"[DEBUG] captioner output: {out!r}")
+        first = out[0] if isinstance(out, list) else out
+        raw = first.get("generated_text") or first.get("text") or str(first)
+        raw = raw.strip()
+        print(f"[DEBUG] raw caption: {raw!r}")
+
+        # 3) 장면 분류
+        cls = scene_classifier(img, candidate_labels=SCENE_LABELS)
+        print(f"[DEBUG] scene_classifier output: {cls!r}")
+        scene = cls["labels"][0]
+
+        # 4) 템플릿 매핑
+        template = TEMPLATES.get(scene, "In this picture, {caption}.")
+        result = template.format(caption=raw)
+        print(f"[DEBUG] Final result: {result}")
+        return result
+
+    except Exception as e:
+        # 에러 메시지와 스택 트레이스 리턴
+        error_msg = f"🔴 Error:\n{e}\n\n{traceback.format_exc()}"
+        print(error_msg)
+        return error_msg