I made an AI system that turns 2D floor plans into 3D, kinda surprised it works

2026-05-29 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning · Depth: Intermediate, short

Summary

An AI system named FloorVis converts 2D floor plans from real estate listings into 3D renders, complete with furniture and lighting. The creator initially encountered "AI hallucinations," where the model would inaccurately alter wall placements, add windows, or change room sizes. This was mitigated by implementing a two-step process: first, the system reads and locks in the actual layout, including walls and doors, and then it proceeds with the 3D rendering. While open floor plans remain a challenge, FloorVis is now online and aims to provide quick visualizations, generating a 3D render in approximately 2 minutes. User feedback, however, highlights that input plans often require significant pre-processing (cropping ancillary information) and that the system can still produce hallucinated rooms or flipped directions, indicating ongoing challenges with consistency and accuracy for varied input quality.

Key takeaway

For AI Engineers developing generative models that must maintain strict fidelity to input data, you should implement multi-stage processing. First, ensure your system accurately parses and locks in critical structural elements from the input, like FloorVis does with walls and doors. Then, apply generative steps for creative elements. This approach helps mitigate "AI hallucinations" and ensures outputs remain consistent with the source, though extensive testing with varied, real-world inputs is crucial to address edge cases and improve reliability.

Key insights

Mitigating "AI hallucinations" in generative models often requires breaking down complex tasks into constrained, sequential steps.

Principles

• Constrain generative AI outputs to preserve input fidelity
• Prioritize input parsing before creative generation

Method

The system first reads and locks in the 2D layout's walls and doors, then generates the 3D elements separately.

In practice

• Implement multi-stage generation for complex image-to-image tasks
• Pre-process inputs to remove extraneous data for better AI parsing

Topics

• 2D to 3D Conversion
• Generative AI
• AI Hallucinations
• Floor Plan Visualization
• Multi-stage Generation
• Real Estate Technology

Best for: Computer Vision Engineer, AI Engineer, Machine Learning Engineer, Entrepreneur

Related on AIssential

• Can These Views Be One Scene? Evaluating Multiview 3D Consistency when 3D Foundation Models Hallucinate — Multiview 3D consistency metrics can fail when 3D foundation models hallucinate, requiring more robust evaluation.
• Perplexity Kills Ads Over Trust Concerns — Achieving photorealistic AI images requires detailed prompting, structured inputs, reprompting from real photos, and post-processing imperfections.
• Generative Floor Plan Design with LLMs via Reinforcement Learning with Verifiable Rewards — LLMs can generate highly constrained, structured floor plans by combining supervised fine-tuning with reinforcement learning and verifiable rewards.
• Reinforcing 3D Understanding in Point-VLMs via Geometric Reward Credit Assignment — Targeted reward assignment and geometric consistency verification significantly improve 3D spatial reasoning in Point-VLMs.
• Code-as-Room: Generating 3D Rooms from Top-Down View Images via Agentic Code Synthesis — Code-as-Room generates 3D rooms from top-down images using MLLM-driven Blender code synthesis with a structured execution harness.
• Why Your AI Assistant Confidently Lies — And Why It’s Not the Data’s Fault — LLM hallucinations stem from three architectural mechanisms, not just bad training data.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.