Update app.py

app.py

@@ -1,79 +1,62 @@
 import gradio as gr
-import torch
 import numpy as np
 from PIL import Image
+
 from inference_utils import create_model, inference
-from cv2_utils import getContours      
 
-# ------------------------------
-# 1) 옵션 클래스 (DeepCrack 기본 설정)
-# ------------------------------
+# --------- DeepCrack 옵션 구성 ---------
 class Opt:
+    # 기본값
+    checkpoints_dir = "./checkpoints"
+    name = "deepcrack"
     gpu_ids = []
     isTrain = False
-    checkpoints_dir = "."
-    model = "deepcrack"
+
     input_nc = 3
-    output_nc = 3
+    num_classes = 1
     ngf = 64
-    netG = "deepcrack"
-    norm = "batch"
-    use_dropout = False
+
+    norm = "instance"
     init_type = "normal"
     init_gain = 0.02
 
-opt = Opt()
+    display_sides = False
+    loss_mode = "bce"
+    lr = 0.001
 
-# ------------------------------
-# 2) 모델 로드
-# ------------------------------
+
+# --------- 모델 로드 ---------
 print("🔥 Loading DeepCrack model...")
+opt = Opt()
 model = create_model(opt, cp_path="pretrained_net_G.pth")
-print("✅ Model loaded successfully!")
-
-# ------------------------------
-# 3) 예측 함수
-# ------------------------------
-def predict(img: Image.Image):
-
-    # PIL → bytes 변환
-    buf = Image.new("RGB", img.size)
-    buf.paste(img)
-    bytes_img = cv2.imencode(".jpg", np.array(buf))[1].tobytes()
-
-    # 추론 실행
-    result_img, visuals = inference(
-        model,
-        bytes_img,
-        dim=img.size,
-        unit="px"
-    )
-
-    # Pillow로 변환하여 출력
-    out_img = Image.fromarray(result_img)
+print("🔥 DeepCrack model loaded!")
 
-    # JSON은 “균열 여부 + 확률”만 전달
-    prob = float(visuals["fused"].max() / 255.0)
 
-    has_crack = prob >= 0.5
-
-    return out_img, {
-        "hasCrack": has_crack,
-        "confidence": round(prob, 4)
+# --------- 예측 함수 ---------
+def predict(img: Image.Image):
+    output_img, confidence = inference(model, img)
+
+    has_crack = confidence > 0.5
+    label = "crack" if has_crack else "normal"
+
+    return {
+        "data": [
+            {
+                "label": label,
+                "confidence": float(confidence)
+            }
+        ]
     }
 
-# ------------------------------
-# 4) Gradio Interface
-# ------------------------------
+
+# --------- Gradio API 인터페이스 ---------
 demo = gr.Interface(
     fn=predict,
     inputs=gr.Image(type="pil"),
-    outputs=[
-        gr.Image(label="Crack Segmentation Output"),
-        gr.JSON(label="Prediction")
-    ],
-    title="DeepCrack Segmentation Model",
-    description="Detects crack regions and generates segmentation overlays."
+    outputs=gr.JSON(label="Detection Result"),
+    title="DeepCrack — Concrete Crack Detection",
+    description="딥러닝 기반 콘크리트 균열 segmentation 모델 DeepCrack",
+    flagging_mode="never"
 )
 
 if __name__ == "__main__":