TU-CSE Dataset

This dataset presents high-quality point cloud data and corresponding 3D models (LOD200) from five distinct residential environments. The dataset is designed to support advancements in as-built building modeling, semantic segmentation, and object detection, addressing the challenges of diverse architectural styles, spatial layouts, and real-world conditions. Each dataset includes raw point cloud data processed to ensure precision and completeness, along with manually reconstructed 3D models conforming to Level of Detail 200 (LOD200). These models provide simplified geometric representations suitable for spatial analysis, topology extraction, and artificial intelligence applications. The five datasets comprise: Multi-Floor Houses (D1, D2), showcasing complex multi-level layouts with cluttered interiors; a House Under Construction (D3), emphasizing the unique characteristics of unfinished structures; a Student Dormitory (D4), representing dynamic, fully furnished shared living spaces captured via MLS; and a High-Rise Apartment (D5), illustrating compact urban living with open floorplans and natural lighting. The inclusion of both raw point cloud data and LOD200 models offers a versatile foundation for exploring a wide range of applications, including automated space classification, topology modeling, and AI-driven object recognition. By making this dataset publicly accessible, the research aims to bridge gaps in current datasets by providing high-quality, diverse data for academic and industrial use. It invites researchers to test algorithms, develop new methodologies, and explore innovative solutions for building assessment and reconstruction tasks. This contribution is expected to stimulate advancements in point cloud processing and as-built modeling, offering valuable insights into the capabilities and limitations of current algorithms while encouraging the development of robust, scalable approaches for real-world applications.