Freehand Sketch Image Generation from 3D Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for generating freehand sketch data for machine learning either require extensive user input, which is time-consuming and costly, or automated solutions that fail to retain the inherent characteristics of user sketches, making it difficult to generalize freehand sketches for machine learning databases.
Innovation Solution
A method and system for automatically generating 2D freehand sketch data from 3D models, involving the creation of multiple sketch images with strokes and guide lines by adjusting vertices, determining projection matrices, and applying thickness, transparency, and texture profiles, to produce training data for machine learning models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If automated methods are used to acquire sketch data, then time and cost consumption is reduced, but the inherent characteristics of user sketches are not retained
Solution Approach 1:
The system creates synthetic sketch data by copying and transforming geometric features from 3D models into 2D sketch representations. Multiple view projections (front, side, top views) are generated and processed to create training data pairs without requiring actual user sketches, thus retaining structural characteristics while automating the data acquisition process
Solution Approach 2:
The system applies parameter transformations including random scaling, rotation, and translation to the generated sketch images. These parameter changes create diverse training samples that maintain the essential geometric characteristics of user sketches while providing sufficient variation for machine learning model generalization
2Device complexity
If single-line freehand sketches are generated from 3D models, then machine learning complexity is simplified, but the sketches do not faithfully reflect designer's actual sketch design
Solution Approach 1:
The system segments the sketch generation process into multiple independent view projections (front view, side view, top view). Each view is processed separately to extract edge maps and generate sketch images, allowing the complex 3D-to-2D transformation to be broken down into manageable components that collectively preserve design characteristics
Solution Approach 2:
The system transforms 3D model data into multiple 2D view projections, creating a multi-dimensional representation approach. By generating sketches from different viewing angles and combining them, the system captures comprehensive design information that single-line sketches miss, while maintaining compatibility with machine learning input requirements
Data Source
AI summary
Provided is a method for generating a freehand sketch of a 3D model for machine learning, which is executed by one or more processors, in which a method for generating freehand sketch data for machine learning includes receiving a 3D model of a target object, and generating a plurality of different freehand sketch images for the target object based on the 3D model.