VisioPlan

Development of an AI for generating 3D-building-models from 2D-data

Description of the company

VisioPlan revolutionizes the construction industry by addressing the persistent issue of planning errors, which are a major source of cost overruns and delays. Despite the obvious benefits of Virtual Reality (VR) in construction, it has not yet become standard practice due to high implementation barriers and time-consuming processes involved. VisioPlan overcomes these challenges by offering a full-service solution that automates 3D modeling,
texturing, and lighting from simple 2D floorplans, making VR accessible and intuitive for both planners and clients. Our software allows clients to virtually walk through their future homes, identify potential issues, and communicate changes in real time. This reduces planning errors, cuts down on personal meetings, and helps avoid the significant costs and environmental impact associated with construction mistakes. VisioPlan's innovative VR
approach creates a decentralised design process that improves collaboration and delivers a seamless, interactive experience that benefits everyone involved in the construction process.

Situation

We have defined a basic structure and requirements for an AI. We have access to relevant data and collect additional data from the floorplan and 3D models we have created. We provide a workspace in the startup port where we can work together. And we are participating in the AI Accelerator of the AI-Hub Hamburg.

Problem

In theory, there should be a high availability of 3D models in residential construction due to the increasing digitalization of the construction industry in recent years. In practice, however, 3D models are still not the norm in residential construction due to the large number of existing buildings and outdated design methods. This means that information is still exchanged on the basis of 2D plans, which leads to a high loss of information and
dissatisfaction in the planning process with the client. VisioPlan takes these 2D plans and prepares them for virtual reality. This preparation is a repetitive and tedious process that is still done manually. The introduction of AI-based processes is expected to increase the efficiency of this preparation process.

Aims of the project

The aim of the Fishing project is to build a prototype of an AI which creates 3D models from
2D plans. VisioPlan will use this prototype to apply for the second phase of the federal
government's funding project.

Scopes:

The scope of the project is to develop an AI that improves the efficiency of the preparation process (2D floorplan to VR-ready 3D model) described above. You will work in an independent team of students with different skills and preferences. The first step is to identify the individual sub-processes that can be taken over by the AI. The required architectures and data sets will be defined by means of an overall concept. In a preparation phase, the existing data sets for the individual sub-processes are viewed and compiled. The concept is then tested by developing a prototype and, if necessary, iterated.

Phase 1: Project Kick-off & Workshops (10 hours per student)
1. Initial Workshop (4 hours):
○ Introduction to VisioPlan’s mission and AI challenges.
○ Overview of current manual 3D model generation processes.
○ Presentation of AI structure and requirements.
○ Q&A session with the CTO.
2. Data Collection & Familiarization (6 hours):
○ Access and review existing data (2D blueprints and 3D models).
○ Group discussions to understand the scope and datasets.
○ Identification of key challenges in automating model generation.

Phase 2: Research & Design of AI Model (25 hours per student)
1. Research on AI Algorithms (10 hours):
○ Each student assigned research topics: AI model structures, machine learning,
and 3D model generation.
○ Weekly brainstorming meetings to share findings and discuss applicability.
2. Design of AI Architecture (15 hours):
○ Collaborative design of AI framework.
○ Outline key components: image processing, feature extraction, conversion
from 2D to 3D, etc.
○ Prepare initial sketches for AI processes.

Phase 3: Prototyping the AI Model (35 hours per student)
1. Data Preprocessing & Setup (10 hours):
○ Organizing collected 2D blueprints and 3D model data.
○ Writing code for data handling and preprocessing for the AI model.
2. AI Development (20 hours):
○ Students work in pairs to build various components of the AI system.
○ Coding algorithms for converting 2D data into 3D outputs.
○ Weekly check-ins with CTO to ensure progress and troubleshoot.
3. Internal Testing of AI (5 hours):
○ Test initial versions of the AI on small sets of blueprints.
○ Document errors and areas for improvement.

Phase 4: Refinement & Final Presentation (20 hours per student)
1. Debugging & Refinement (10 hours):
○ Debugging issues found during testing.
○ Optimization of AI model for accuracy and speed.
2. Final Workshop & Presentation Preparation (10 hours):
○ Preparation of prototype demonstration.
○ Workshop to refine the presentation of the results.
○ Presentation to VisioPlan’s management team, discussing learnings and next
steps.

Target group (students)

We are searching for young, passionate talents who are eager to make a real impact. Your  grades aren't what matter most to us—it your enthusiasm and dedication to the project that count. We're assembling a team capable of developing groundbreaking AI, and if you believe you can make a decisive contribution, we want to hear from you! Our CTO will guide the process, ensuring the project stays on course and providing the foundational structure we need to succeed.

We hold ourselves to the highest standards in everything we do, and we expect the same from this project. Mistakes are not just inevitable; they are essential. They are opportunities to learn, to improve, and to grow. This project will be bigger than any of us as individuals—it will challenge us to rise above and go beyond. Just like founding a start-up, great things can only be achieved when we push past our limits. This is your chance to be part of something extraordinary, your chance to grow and create something truly impactful.

Have you already built an AI and are you familiar with common AI frameworks? Do you also speak Python? Would you also like to work in an ambitious start-up? Then get in touch, because we are ready to create something new together.

Dates
Please save these dates: Fishing for Experience Termine

Registration
You can apply for Fishing for Experience online.