Update app.py (#1)

- Update app.py (1b807a05f70700e1b5f655a7671e85c2dd947230)


Co-authored-by: Zheheng Li <Martyr1@users.noreply.huggingface.co>

app.py

@@ -1,4 +1,3 @@
-import os
 import cv2
 import numpy as np
 import torch
@@ -9,36 +8,37 @@ import boto3
 import uuid
 import io
 from glob import glob
+import os
 from pipeline.ImgOutlier import detect_outliers
 from pipeline.normalization import align_images
 
-# 检测是否在Hugging Face环境中运行
+# Detect if running inside Hugging Face Spaces
 HF_SPACE = os.environ.get('SPACE_ID') is not None
 
-# DigitalOcean Spaces上传函数
+# DigitalOcean Spaces upload function
 def upload_mask(image, prefix="mask"):
     """
-    将分割掩码图像上传到DigitalOcean Spaces
-    
+    Upload segmentation mask image to DigitalOcean Spaces
+
     Args:
-        image: PIL Image对象
-        prefix: 文件名前缀
+        image: PIL Image object
+        prefix: filename prefix
         
     Returns:
-        上传文件的URL
+        Public URL of the uploaded file
     """
     try:
-        # 从环境变量获取凭据
+        # Get credentials from environment variables
         do_key = os.environ.get('DO_SPACES_KEY')
         do_secret = os.environ.get('DO_SPACES_SECRET')
         do_region = os.environ.get('DO_SPACES_REGION')
         do_bucket = os.environ.get('DO_SPACES_BUCKET')
         
-        # 校验凭据是否存在
+        # Check if credentials exist
         if not all([do_key, do_secret, do_region, do_bucket]):
-            return "DigitalOcean凭据未设置"
+            return "DigitalOcean credentials not set"
         
-        # 创建S3客户端
+        # Create S3 client
         session = boto3.session.Session()
         client = session.client('s3',
                                region_name=do_region,
@@ -46,15 +46,15 @@ def upload_mask(image, prefix="mask"):
                                aws_access_key_id=do_key,
                                aws_secret_access_key=do_secret)
         
-        # 生成唯一文件名
+        # Generate unique filename
         filename = f"{prefix}_{uuid.uuid4().hex}.png"
         
-        # 将图像转换为字节流
+        # Convert image to bytes
         img_byte_arr = io.BytesIO()
         image.save(img_byte_arr, format='PNG')
         img_byte_arr.seek(0)
         
-        # 上传到Spaces
+        # Upload to Spaces
         client.upload_fileobj(
             img_byte_arr,
             do_bucket,
@@ -62,13 +62,13 @@ def upload_mask(image, prefix="mask"):
             ExtraArgs={'ACL': 'public-read', 'ContentType': 'image/png'}
         )
         
-        # 返回公共URL
+        # Return public URL
         url = f'https://{do_bucket}.{do_region}.digitaloceanspaces.com/{filename}'
         return url
     
     except Exception as e:
-        print(f"上传失败: {str(e)}")
-        return f"上传错误: {str(e)}"
+        print(f"Upload failed: {str(e)}")
+        return f"Upload error: {str(e)}"
 
 # Global Configuration
 MODEL_PATHS = {
@@ -101,11 +101,11 @@ COLORS = [
 # Load model function
 def load_model(model_path, device="cuda"):
     try:
-        # 如果在HF环境中，默认使用CPU
+        # If running inside HF Spaces, default to CPU
         if HF_SPACE:
-            device = "cpu"  # HF Space可能没有GPU
+            device = "cpu"
         elif not torch.cuda.is_available():
-            device = "cpu"  # 本地环境也可能没有GPU
+            device = "cpu"
         
         model = smp.create_model(
             "DeepLabV3Plus",
@@ -120,30 +120,30 @@ def load_model(model_path, device="cuda"):
         model.load_state_dict(state_dict)
         model.to(device)
         model.eval()
-        print(f"模型加载成功: {model_path}")
+        print(f"Model loaded successfully: {model_path}")
         return model
     except Exception as e:
-        print(f"模型加载失败: {e}")
+        print(f"Model loading failed: {e}")
         return None
 
 # Load reference vector
 def load_reference_vector(vector_path):
     try:
         if not os.path.exists(vector_path):
-            print(f"参考向量文件不存在: {vector_path}")
+            print(f"Reference vector file not found: {vector_path}")
             return []
         ref_vector = np.load(vector_path)
-        print(f"参考向量加载成功: {vector_path}")
+        print(f"Reference vector loaded successfully: {vector_path}")
         return ref_vector
     except Exception as e:
-        print(f"参考向量加载失败 {vector_path}: {e}")
+        print(f"Reference vector loading failed {vector_path}: {e}")
         return []
 
-# Load reference image
+# Load reference images
 def load_reference_images(ref_dir):
     try:
         if not os.path.exists(ref_dir):
-            print(f"参考图像目录不存在: {ref_dir}")
+            print(f"Reference image directory not found: {ref_dir}")
             os.makedirs(ref_dir, exist_ok=True)
             return []
             
@@ -157,10 +157,10 @@ def load_reference_images(ref_dir):
             img = cv2.imread(file)
             if img is not None:
                 reference_images.append(img)
-        print(f"从 {ref_dir} 加载了 {len(reference_images)} 张图像")
+        print(f"Loaded {len(reference_images)} images from {ref_dir}")
         return reference_images
     except Exception as e:
-        print(f"加载图像失败 {ref_dir}: {e}")
+        print(f"Image loading failed {ref_dir}: {e}")
         return []
 
 # Preprocess the image
@@ -208,7 +208,6 @@ def create_overlay(image, segmentation_map, alpha=0.5):
     if image.shape[:2] != segmentation_map.shape[:2]:
         segmentation_map = cv2.resize(segmentation_map, (image.shape[1], image.shape[0]), interpolation=cv2.INTER_NEAREST)
     return cv2.addWeighted(image, 1-alpha, segmentation_map, alpha, 0)
-
 # Perform segmentation
 def perform_segmentation(model, image_bgr):
     device = "cuda" if torch.cuda.is_available() and not HF_SPACE else "cpu"
@@ -225,18 +224,18 @@ def perform_segmentation(model, image_bgr):
 # Single image processing
 def process_coastal_image(location, input_image):
     if input_image is None:
-        return None, None, "请上传一张图片", "未检测", None
+        return None, None, "Please upload an image", "Not detected", None
     
     device = "cuda" if torch.cuda.is_available() and not HF_SPACE else "cpu"
     model = load_model(MODEL_PATHS[location], device)
     
     if model is None:
-        return None, None, f"错误：无法加载模型", "未检测", None
+        return None, None, f"Error: Failed to load model", "Not detected", None
     
     ref_vector = load_reference_vector(REFERENCE_VECTOR_PATHS[location])
     ref_images = load_reference_images(REFERENCE_IMAGE_DIRS[location])
     
-    outlier_status = "未检测"
+    outlier_status = "Not detected"
     is_outlier = False
     image_bgr = cv2.cvtColor(np.array(input_image), cv2.COLOR_RGB2BGR)
     
@@ -247,35 +246,35 @@ def process_coastal_image(location, input_image):
         filtered, _ = detect_outliers(ref_images, [image_bgr])
         is_outlier = len(filtered) == 0
     else:
-        print("警告：没有参考图像或参考向量可用于异常检测")
+        print("Warning: No reference images or reference vectors available for outlier detection")
         is_outlier = False
     
-    outlier_status = "异常检测: <span style='color:red;font-weight:bold'>未通过</span>" if is_outlier else "异常检测: <span style='color:green;font-weight:bold'>通过</span>"
+    outlier_status = "Outlier Detection: <span style='color:red;font-weight:bold'>Failed</span>" if is_outlier else "Outlier Detection: <span style='color:green;font-weight:bold'>Passed</span>"
     seg_map, overlay, analysis = perform_segmentation(model, image_bgr)
     
-    # 尝试上传到DigitalOcean Spaces
-    url = "本地存储"
+    # Try uploading to DigitalOcean Spaces
+    url = "Local Storage"
     try:
         url = upload_mask(Image.fromarray(seg_map), prefix=location.replace(' ', '_'))
     except Exception as e:
-        print(f"上传失败: {e}")
-        url = f"上传错误: {str(e)}"
+        print(f"Upload failed: {e}")
+        url = f"Upload error: {str(e)}"
     
     if is_outlier:
-        analysis = "<div style='color:red;font-weight:bold;margin-bottom:10px'>警告：图像未通过异常检测，结果可能不准确！</div>" + analysis
+        analysis = "<div style='color:red;font-weight:bold;margin-bottom:10px'>Warning: The image failed outlier detection, the result may be inaccurate!</div>" + analysis
     
     return seg_map, overlay, analysis, outlier_status, url
 
-# Spacial Alignment
+# Spatial Alignment
 def process_with_alignment(location, reference_image, input_image):
     if reference_image is None or input_image is None:
-        return None, None, None, None, "请上传参考图像和需要分析的图像", "未处理", None
+        return None, None, None, None, "Please upload both reference and target images", "Not processed", None
     
     device = "cuda" if torch.cuda.is_available() and not HF_SPACE else "cpu"
     model = load_model(MODEL_PATHS[location], device)
     
     if model is None:
-        return None, None, None, None, "错误：无法加载模型", "未处理", None
+        return None, None, None, None, "Error: Failed to load model", "Not processed", None
     
     ref_bgr = cv2.cvtColor(np.array(reference_image), cv2.COLOR_RGB2BGR)
     tgt_bgr = cv2.cvtColor(np.array(input_image), cv2.COLOR_RGB2BGR)
@@ -284,20 +283,20 @@ def process_with_alignment(location, reference_image, input_image):
         aligned, _ = align_images([ref_bgr, tgt_bgr], [np.zeros_like(ref_bgr), np.zeros_like(tgt_bgr)])
         aligned_tgt_bgr = aligned[1]
     except Exception as e:
-        print(f"空间对齐失败: {e}")
-        return None, None, None, None, f"空间对齐失败: {str(e)}", "处理失败", None
+        print(f"Spatial alignment failed: {e}")
+        return None, None, None, None, f"Spatial alignment failed: {str(e)}", "Processing failed", None
     
     seg_map, overlay, analysis = perform_segmentation(model, aligned_tgt_bgr)
     
-    # 尝试上传到DigitalOcean Spaces
-    url = "本地存储"
+    # Try uploading to DigitalOcean Spaces
+    url = "Local Storage"
     try:
         url = upload_mask(Image.fromarray(seg_map), prefix="aligned_" + location.replace(' ', '_'))
     except Exception as e:
-        print(f"上传失败: {e}")
-        url = f"上传错误: {str(e)}"
+        print(f"Upload failed: {e}")
+        url = f"Upload error: {str(e)}"
     
-    status = "空间对齐: <span style='color:green;font-weight:bold'>完成</span>"
+    status = "Spatial Alignment: <span style='color:green;font-weight:bold'>Completed</span>"
     ref_rgb = cv2.cvtColor(ref_bgr, cv2.COLOR_BGR2RGB)
     aligned_tgt_rgb = cv2.cvtColor(aligned_tgt_bgr, cv2.COLOR_BGR2RGB)
     
@@ -305,61 +304,59 @@ def process_with_alignment(location, reference_image, input_image):
 
 # Create the Gradio interface
 def create_interface():
-    # 设置统一的显示尺寸
-    disp_w, disp_h = 683, 512  # 统一设置宽高比
-    
-    with gr.Blocks(title="海岸侵蚀分析系统") as demo:
-        gr.Markdown("""# 海岸侵蚀分析系统
+    # Set unified display size
+    disp_w, disp_h = 683, 512  # Maintain aspect ratio
+
+    with gr.Blocks(title="Coastal Erosion Analysis System") as demo:
+        gr.Markdown("""# Coastal Erosion Analysis System
 
-上传海岸照片进行分析，包括分割和空间对齐功能。""")
+Upload coastal images for analysis, including segmentation and spatial alignment.""")
         with gr.Tabs():
-            with gr.TabItem("单张图像分割"):
+            with gr.TabItem("Single Image Segmentation"):
                 with gr.Row():
-                    loc1 = gr.Radio(list(MODEL_PATHS.keys()), label="选择模型", value=list(MODEL_PATHS.keys())[0])
+                    loc1 = gr.Radio(list(MODEL_PATHS.keys()), label="Select Model", value=list(MODEL_PATHS.keys())[0])
                 with gr.Row():
-                    # 确保所有图像组件使用相同的尺寸
-                    inp = gr.Image(label="输入图像", type="numpy", image_mode="RGB", height=disp_h, width=disp_w)
-                    seg = gr.Image(label="分割图像", type="numpy", height=disp_h, width=disp_w)
-                    ovl = gr.Image(label="叠加图像", type="numpy", height=disp_h, width=disp_w)
+                    inp = gr.Image(label="Input Image", type="numpy", image_mode="RGB", height=disp_h, width=disp_w)
+                    seg = gr.Image(label="Segmentation Map", type="numpy", height=disp_h, width=disp_w)
+                    ovl = gr.Image(label="Overlay Image", type="numpy", height=disp_h, width=disp_w)
                 with gr.Row():
-                    btn1 = gr.Button("执行分割")
-                url1 = gr.Text(label="分割图URL")
-                status1 = gr.HTML(label="异常检测状态")
-                res1 = gr.HTML(label="分析结果")
+                    btn1 = gr.Button("Run Segmentation")
+                url1 = gr.Text(label="Segmentation Image URL")
+                status1 = gr.HTML(label="Outlier Detection Status")
+                res1 = gr.HTML(label="Analysis Result")
                 btn1.click(fn=process_coastal_image, inputs=[loc1, inp], outputs=[seg, ovl, res1, status1, url1])
             
-            with gr.TabItem("空间对齐分割"):
+            with gr.TabItem("Spatial Alignment Segmentation"):
                 with gr.Row():
-                    loc2 = gr.Radio(list(MODEL_PATHS.keys()), label="选择模型", value=list(MODEL_PATHS.keys())[0])
+                    loc2 = gr.Radio(list(MODEL_PATHS.keys()), label="Select Model", value=list(MODEL_PATHS.keys())[0])
                 with gr.Row():
-                    # 确保所有图像组件使用相同的尺寸
-                    ref_img = gr.Image(label="参考图像", type="numpy", image_mode="RGB", height=disp_h, width=disp_w)
-                    tgt_img = gr.Image(label="待分析图像", type="numpy", image_mode="RGB", height=disp_h, width=disp_w)
+                    ref_img = gr.Image(label="Reference Image", type="numpy", image_mode="RGB", height=disp_h, width=disp_w)
+                    tgt_img = gr.Image(label="Target Image", type="numpy", image_mode="RGB", height=disp_h, width=disp_w)
                 with gr.Row():
-                    btn2 = gr.Button("执行空间对齐分割")
+                    btn2 = gr.Button("Run Spatial Alignment and Segmentation")
                 with gr.Row():
-                    orig = gr.Image(label="原始图像", type="numpy", height=disp_h, width=disp_w)
-                    aligned = gr.Image(label="对齐后图像", type="numpy", height=disp_h, width=disp_w)
+                    orig = gr.Image(label="Original Image", type="numpy", height=disp_h, width=disp_w)
+                    aligned = gr.Image(label="Aligned Image", type="numpy", height=disp_h, width=disp_w)
                 with gr.Row():
-                    seg2 = gr.Image(label="分割图像", type="numpy", height=disp_h, width=disp_w)
-                    ovl2 = gr.Image(label="叠加图像", type="numpy", height=disp_h, width=disp_w)
-                url2 = gr.Text(label="分割图URL")
-                status2 = gr.HTML(label="空间对齐状���")
-                res2 = gr.HTML(label="分析结果")
+                    seg2 = gr.Image(label="Segmentation Map", type="numpy", height=disp_h, width=disp_w)
+                    ovl2 = gr.Image(label="Overlay Image", type="numpy", height=disp_h, width=disp_w)
+                url2 = gr.Text(label="Segmentation Image URL")
+                status2 = gr.HTML(label="Alignment Status")
+                res2 = gr.HTML(label="Analysis Result")
                 btn2.click(fn=process_with_alignment, inputs=[loc2, ref_img, tgt_img], outputs=[orig, aligned, seg2, ovl2, res2, status2, url2])
     return demo
 
 if __name__ == "__main__":
-    # 创建必要的目录
+    # Create necessary directories
     for path in ["models", "reference_images/MM", "reference_images/SJ"]:
         os.makedirs(path, exist_ok=True)
     
-    # 检查模型文件是否存在
+    # Check if model files exist
     for p in MODEL_PATHS.values():
         if not os.path.exists(p):
-            print(f"警告：模型文件 {p} 不存在！")
+            print(f"Warning: Model file {p} does not exist!")
     
-    # 检查DigitalOcean凭据是否存在
+    # Check if DigitalOcean credentials exist
     do_creds = [
         os.environ.get('DO_SPACES_KEY'),
         os.environ.get('DO_SPACES_SECRET'),
@@ -367,11 +364,10 @@ if __name__ == "__main__":
         os.environ.get('DO_SPACES_BUCKET')
     ]
     if not all(do_creds):
-        print("警告：DigitalOcean Spaces凭据不完整，上传功能可能不可用")
+        print("Warning: Incomplete DigitalOcean Spaces credentials, upload functionality may not work")
     
-    # 创建并启动界面
+    # Create and launch the interface
     demo = create_interface()
-    # 在HF环境中使用适当的启动配置
     if HF_SPACE:
         demo.launch()
     else: