Langhui
DataAssetsAPI . Engineering data products

CAD
Project drawing dataset

Covers 5,000+ sets of CAD drawings for construction engineering and mechanical design. This covers a variety of specialized drawings, including master plans, building plans, structural construction plans, equipment construction plans, electrical maps, etc., and provides DWG/DXF original format and structured label data, which are specially designed for AI smart review maps, drawing identification and large model training scenarios.

5,000+ sets of drawings DWG + DXF Dual Formats StructuredAnnotation AI audit chart ready.

6 main categories

Professional drawing type

100%

The certification was approved.

10

Time spans

2

DWG + DXF Dual Formats

Dataset Overview

A comprehensive understanding of the composition and core features of this data set provides a high-quality engineering drawing data base for AI review maps, smart identification and large model training.

The core building plans, plans, profiles, detailed drawings, etc. cover a wide variety of types of housing, commerce, public buildings, etc.

(c) Structural drawings, such as basic maps, column charts, beryllium charts, sheet charts, stairwells, etc., containing steel-concrete and steel structures.

Specialized construction drawings for equipment such as drainage, HAC, fire protection, including system maps, floor layout and detailed maps.

Electrical design drawings for distribution systems, lighting floors, weak electrical systems, mine-protected terrain, etc.

Data Dimensions and Distribution

The data sets are carefully screened and matched to ensure that the various types of drawings cover a comprehensive and balanced distribution and meet the needs of diverse model training.

Type distribution of drawings

Percentage of professional type

25%
20%
20%
15%
General and mechanical
20%

Building type coverage

Housing

35%

Commercial

25%

Public buildings

20%

Industry

15%

Other

5%

Figure Complexity Distribution

Simple (1-2 floor)
15%
Medium (3-5 floor)
25%
Complex (6-10 floors)
40%
Super complicated (>10 floors)
20%

Data quality characteristics

Format standards: DWG (AutoCAD 2013+) and DXF common exchange format double-versions available
Figure integrity: Each set of drawings contains a full professional package, layered and marked
Annotation QualityComponent recognition, space labels, dimensions labels manually verified, 95% accuracy
Diagram Information: complete maintenance of graphic metadata information such as layers, lines, blocks, text
Engineering phase:: Cover the phases of the programme design, preliminary design, construction design
Copyright Clarity: All drawings are legal, copyright is clear, authorization is retroactive

Structured Mark Overrides

Space recognition

Rooms/regions/functional partitions

Component Recognition

Walls/posts/ beams/doors/windows

Text Comment

Size/higher/state text

Layer Classification

Layer Semantic Tab Map

Licensed Application Scenarios

This data set is legally authorized and can be used in the following three core scenarios. All data have been copyrighted and quality validated.

AI Smart View

Provides large-scale training data for the AI smart-reviewed mapping system, covering scenarios such as architectural specifications verification, building recognition, space compliance testing and supporting automated graphic review.

  • Automatic verification of normative provisions
  • Component Smart Identification Positioning
  • Auto-scoring of the quality of the drawing

Figure recognition and digitization

Supporting the intellectual resolution and digitization of CAD drawings, achieving the image recognition, text extraction, spatial relationship understanding, serving BIM reverse modelling and digitization of drawing archives.

  • Drawing of text from the drawing OCR
  • BIM Reverse Modelling Support
  • Map structured information extraction

Large model training and academic research

Provide high-quality CAD drawing training data for large-language and multi-modular models in engineering, and support the mapping understanding, design generation, and specification of forward research directions such as questions and answers.

  • Engineering MIM pre-training
  • AI-assisted design generation
  • Engineering knowledge spectroscopy

Licensing and Compliance Assurance

Data are legally sourced, tenure is clear and intellectual property and industry norms are strictly adhered to.

Basic Licensing Information

Dataset Name CAD Project Map Data Set
Nature of License Non-Exclusive License
License Term May 6, 2026 — May 1, 2027
Data Holder Cooperative Design Institutes Alliance
Operator Changsha Langhui Information Technology Co., Ltd.
Scope of License Acquisition, De-identified Annotation, Structuring, Storage and Use

Compliance and Security Assurance

Intellectual property compliance

All drawings are copyrighted clearly and legally authorized to strictly comply with the provisions of the Copyright Act

Legal Compliance Assurance

Strict adherence to the Data Security Act, the Personal Information Protection Act and the code of engineering industry

Sublicensing Prohibited

The data must not be sublicensed, resold, disclosed or shared with any unauthorized third party

Clear Ownership

The licensor guarantees that the data sources are lawful, ownership is clear and free of defects in title, and that legal usage rights have been obtained