Created on 26th October 2024
•
Designers, engineers, and architects often require digital representations of physical objects or spaces in CAD and 3D formats for efficient planning, visualization, and editing. Currently, creating these models involves manual design, which is time-consuming and may lack precision. This AI solution automates the process, converting real-life images of structures, vehicles, and machinery into editable AutoCAD (2D) and Blender (3D) files based on user prompts, significantly saving time and enhancing workflow precision.
Data Filtering & Quality Control: Preprocessing the dataset was crucial to remove irrelevant information and ensure clean data, given the diversity of images and annotations.
Accurate Entity Extraction: Designing an NLP model that accurately identifies relevant entities (like rooms, dimensions) from textual prompts posed a unique challenge.
3D Model Accuracy: Ensuring realistic 3D outputs in Blender that match the user's specifications and image inputs required tuning model architecture.
AutoCAD Layout Precision: Converting complex prompts and images into structured 2D layouts in AutoCAD necessitated precise geometry generation and error handling.
UI-Model Integration: Creating a smooth interface for users to edit or download models while managing cross-platform compatibility was challenging, especially in Windows and WSL environments.
Tracks Applied (1)
OnDemand by Airev