Procedural rule creator

**FacadeElements – A Dataset of Building Facades with Annotated Windows, Doors, and Balconies**

This dataset contains **3,494 photographs** of real-world building facades, carefully annotated for object detection and instance segmentation tasks.

**Annotated classes:**- **Window**- **Single door** (includes wooden, metal, glass, and other materials)- **Double door**- **Balcony** 

Every image has been manually labeled with **bounding boxes**, making it ideal for training and evaluating deep learning models in architectural element detection, urban scene understanding, and façade parsing.

**Key features:**- 3,494 real-world façade images from diverse architectural styles and regions- Consistent, high-quality annotations following a unified labeling protocol- Doors are categorized as single or double (material variations such as wood, metal, or glass are preserved within the same class)- Suitable for YOLO, Faster R-CNN, Mask R-CNN, Detectron2, MMDetection, and other modern detectors- Ready-to-use for training, validation, and benchmarking

Intended use: computer vision research, automated architectural analysis, building information modeling (BIM), urban documentation, and heritage preservation studies.

Feel free to cite this dataset in your research papers and projects!

**立面元素（FacadeElements）数据集——带标注门窗与阳台的建筑立面数据集**

本数据集包含**3494张实拍建筑立面照片**，针对目标检测与实例分割任务完成了精细标注。

**标注类别：**
- **窗户（Window）**
- **单开门（Single door）**：涵盖木质、金属、玻璃等多种材质
- **双开门（Double door）**
- **阳台（Balcony）**

所有图像均通过**边界框（bounding boxes）**完成人工标注，可完美适配建筑元素检测、城市场景理解及立面解析任务中的深度学习模型训练与评估工作。

**核心特性：**
- 收录来自多元建筑风格与地域的3494张实拍建筑立面图像
- 遵循统一标注规范，具备一致且高质量的标注质量
- 门分为单开门与双开门两类，同一类别内保留木质、金属、玻璃等材质差异
- 适配YOLO、Faster R-CNN、Mask R-CNN、Detectron2、MMDetection等多款现代目标检测器
- 可直接用于模型训练、验证与基准测试

**适用场景：**
计算机视觉研究、自动化建筑分析、建筑信息建模（BIM）、城市档案记录及文物保护研究。

欢迎在研究论文与项目中引用本数据集！