3D Primitive Inference from Partial 2D Sketches
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for converting 2D sketches to 3D models are cumbersome and time-consuming, especially for non-expert users, as they require complete input strokes and lack real-time feedback, making it difficult to achieve accurate 3D model representation from 2D sketches.
Innovation Solution
A computer-implemented method that uses a Convolutional Neural Network to infer 3D primitives from partial 2D sketches, providing real-time feedback by orienting and positioning the 3D primitives to match the view, allowing for progressive sketching and immediate visualization of the 3D model alongside the 2D sketch, using techniques such as bounding squares, rectangles, and binary masks for fitting the 3D projections onto the 2D sketch.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If complete input strokes are required for 3D model inference, then manufacturing precision is improved, but productivity deteriorates
Solution Approach 1:
The system performs 3D model inference using only partial sketch strokes instead of requiring complete input. The neural network processes incomplete sketches and generates 3D primitives progressively, allowing users to see results before finishing their sketch, thereby improving design speed while maintaining acceptable accuracy
Solution Approach 2:
The system performs preliminary 3D model inference during the sketching process itself, rather than waiting for completion. The neural network continuously processes partial strokes and provides progressive feedback, enabling users to validate and adjust their design intent before finalizing the sketch
2Measurement precision
If complete sketch strokes are required before inference, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The neural network performs continuous inference processing throughout the sketching process, rather than waiting for completion. Each partial stroke triggers immediate processing and feedback, eliminating idle waiting time while maintaining precision through progressive refinement of the 3D model as more strokes are added
3Device complexity
If 3D model is inferred from a single 2D sketch, then device complexity is reduced, but measurement precision deteriorates
Solution Approach 1:
The system resolves viewpoint ambiguity by inferring 3D primitives with spatial orientation and positioning that add dimensional context. The neural network determines not just shape but also the viewing angle and perspective from which the 2D sketch was created, effectively adding dimensional information to resolve the ambiguity of single-view sketches
4Speed
If 2D sketch is continuously cropped and resized, then processing speed is improved, but loss of information increases
Solution Approach 1:
The system segments the 2D sketch into multiple regions of interest and processes them separately at different resolutions. Important features are preserved at high resolution while less critical areas are processed at lower resolution, maintaining both processing speed and essential sketch details without uniform cropping and resizing
Data Source
AI summary
A computer-implemented method for designing at least one 3D model in a 3D scene including receiving a user's 2D sketch and displaying it on a plane, said 2D sketch representing a view of the 3D model to be designed, inferring a 3D primitive based on said 2D sketch, said 3D primitive being oriented and positioned in the 3D scene to match said view, performing a 2D projection of the 3D primitive on said plane, and fitting said 2D projection onto said 2D sketch.