3D Shape Stylized Stroke Generation With Editable Vector Curves
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional techniques for generating stylized line drawings from 3D shapes fail to accurately capture the 3D geometry and artist's style, resulting in loss of details and limited editability, especially when converting pixel images.
Innovation Solution
A machine learning model trained using a 2D drawing of a 3D object generates vectorized curves that mimic an artist's style by extracting 3D geometry and applying stroke attributes like thickness and displacement, producing scalable and editable vector graphics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional techniques are used to generate stylized line drawings from 3D shapes, then the generation process is simple, but the accuracy of capturing 3D geometry and artist's style deteriorates, resulting in loss of details
Solution Approach 1:
The system segments the stylized drawing generation into multiple independent neural network components: a first neural network extracts surface geometry features from the 3D shape, a second neural network predicts stroke attributes based on geometry features and style, and a third neural network generates the final stylized drawing. This segmentation allows each component to specialize in one aspect of the transformation, improving overall accuracy while maintaining manageable complexity through modular architecture.
Solution Approach 2:
The system transforms the problem from 2D image processing to 3D geometry processing by extracting surface geometry features (normal vectors, curvature, depth) from the 3D shape. This dimensional elevation enables the model to capture true 3D geometric information rather than relying on 2D image cues, significantly improving the accuracy of geometry capture and style transfer.
2Ease of operation
If pixel images are converted to stylized drawings, then the process is straightforward, but the editability and scalability of the output deteriorates
Solution Approach 1:
The system replaces traditional raster image processing with vector-based curve representation. Instead of manipulating pixel data, the system generates stylized drawings as parametric curves that can be scaled indefinitely without resolution loss. This substitution of mechanical representation (vector vs. raster) simultaneously improves both scalability and editability while maintaining output quality.
3Reliability
If a single neural network is used for stylized drawing generation, then the system is simple, but the ability to preserve fine details and artist's style deteriorates
Solution Approach 1:
The system divides the stylized drawing generation task across three specialized neural networks: the first network extracts surface geometry features (normals, curvature, depth) from the 3D shape, the second network predicts stroke attributes (thickness, displacement, curvature) based on geometry and style, and the third network generates the final stylized drawing. This segmentation enables each network to optimize for its specific function, preserving fine details through specialized feature extraction and improving style fidelity through dedicated attribute prediction.
Solution Approach 2:
The system applies local quality by predicting stroke attributes (thickness, displacement, curvature) at different locations along the curves based on local surface geometry features. Each stroke's properties are determined by the underlying 3D surface characteristics at that specific location, allowing fine details and stylistic variations to be preserved locally rather than applying uniform styling across the entire drawing.
Data Source
AI summary
Techniques for generating a stylized drawing of three-dimensional (3D) shapes using neural networks are disclosed. A processing device generates a set of vector curve paths from a viewpoint of a 3D shape; extracts, using a first neural network of a plurality of neural networks of a machine learning model, surface geometry features of the 3D shape based on geometric properties of surface points of the 3D shape; determines, using a second neural network of the plurality of neural networks of the machine learning model, a set of at least one predicted stroke attribute based on the surface geometry features and a predetermined drawing style; generates, based on the at least one predicted stroke attribute, a set of vector stroke paths corresponding to the set of vector curve paths; and outputs a two-dimensional (2D) stylized stroke drawing of the 3D shape based at least on the set of vector stroke paths.