AI Design Phase Recognition for Sketch Ideation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional product design processes are inefficient and time-consuming, particularly during the ideation phase, where designers must manually generate and refine sketches across multiple design phases, often constrained by individual creativity and expertise.
Innovation Solution
A machine-learning based design system that recognizes the design phase of an initial sketch and generates concept ideas with attributes tailored to that phase, using a multi-part approach involving text prompts, active sketches, and control parameters to automate the ideation process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If designers manually generate and refine sketches across multiple design phases, then design quality and creativity can be maintained, but the process becomes time-consuming and inefficient
Solution Approach 1:
The system enables self-service by automatically generating design sketches and variations without requiring manual intervention from designers. The AI model takes a seed sketch and automatically produces multiple design alternatives across different phases (conceptualization, exploration, refinement), allowing the system to serve itself in the creative process rather than relying on human designers to manually create each iteration.
Solution Approach 2:
The patent replaces the mechanical system of manual sketching with an AI-based computational system. Instead of designers physically drawing and refining sketches by hand through multiple phases, the system uses machine learning models to automatically generate, iterate, and refine design sketches, substituting human manual effort with automated intelligent processes while maintaining design quality.
2Manufacturing precision
If designers manually iterate through multiple design phases, then design refinement can be achieved, but the process requires significant manual effort and repetition
Solution Approach 1:
The system performs preliminary action by pre-defining the design phases (conceptualization, exploration, refinement) and preparing appropriate generation parameters for each phase before actual design work begins. The AI model is pre-configured with phase-specific control parameters that guide the generation process, so when a design task starts, the system already has the refinement strategy ready rather than requiring manual setup for each phase transition.
Solution Approach 2:
The system implements dynamics by making the design generation process adaptive and phase-aware. The AI model dynamically adjusts its generation behavior based on the current design phase, transitioning from broad conceptual exploration to detailed refinement automatically. Control parameters are dynamically modified according to the phase, allowing the system to ease into detailed work progressively rather than requiring manual intervention to change the approach at each phase boundary.
3Measurement precision
If AI systems focus on structured inputs like CAD models, then processing accuracy can be maintained, but early-stage sketches are overlooked despite their importance in ideation
Solution Approach 1:
The system applies parameter changes by transforming the control parameters based on the identified design phase. When processing sketches, the system modifies generation parameters such as semantic similarity thresholds, analogical similarity weights, and variability controls to match the appropriate phase (conceptualization vs. refinement). This allows the AI to maintain high processing accuracy while adapting to the less structured nature of early-stage sketches, as the parameters are tuned to handle the specific characteristics of each phase's input quality.
Solution Approach 2:
The patent implements universality by creating a single AI system that can handle multiple input types (structured CAD models and unstructured sketches) and multiple design phases through a unified framework. The system uses phase identification to determine the appropriate processing mode, making it versatile enough to work with both highly structured and loosely defined inputs while maintaining accuracy. This multi-functional approach eliminates the need for separate systems for different input types.
4Adaptability or versatility
If designers are constrained by individual creativity and expertise, then design consistency can be maintained, but the variety of design alternatives is limited
Solution Approach 1:
The system applies segmentation by breaking down the design generation process into distinct phases (conceptualization, exploration, refinement), each handled by specialized AI processing. Instead of attempting to generate all design alternatives in a single complex operation, the system segments the task into phase-specific generation steps, where each phase produces a subset of alternatives with appropriate characteristics. This reduces overall system complexity by making each segment's function more focused and manageable.
Data Source
AI summary
Systems, methods, and other embodiments described herein relate to improving product ideation through design phase recognition and concept generation. In one embodiment, a method includes acquiring design inputs including at least a text prompt and an active sketch. The method includes, responsive to determining an activity type associated with the active sketch, adapting the design inputs to correspond with the activity type. The method includes generating a guidance image according to the design inputs.