[Chroma]Morph

A Mesh Agent System Based on Color Similarity For Image Style Transfer

2023

Keywords
Computational Design
Agent-based Modeling
Generative Art
Tools
Rhino+Grasshopper
Python Scripting
Advisor
Tobias Schwinn
Teammates
Joyce Wu
My Role
Concept Development
Design Computation
Innovations
Agent-Based Style Transfer: Introduces an agent-based algorithm to transform image styles by having agents influence nearby colors, mimicking wet-on-wet painting.

Localized Transformations: Creates dynamic, localized image changes within a mesh environment, highlighting the interactions between agents and their surroundings.
Similarity-based Agent Model:
Mesh Environment
In our mesh environment, each mesh face acts as an agent, initially in a passive state. Each mesh face has an "active threshold" value that tracks the number of active agents that have passed by, which triggers the creation of new agents once the threshold is reached.

‍
Similarity-based Agent Model:
Agents & Behaviors
Passive & Active Agents
Each active mesh agent seeks out another agent with a similar color and moves toward it. As it moves, the agent influences other mesh faces it encounters, transferring its color at a user-defined decay rate along the path.
‍
Goal:
- Look for and move towards the agents of similar colors
- Influence the agents of the different colors
‍
Attributes:
- State
- Color
- Position
- Neighbors
‍
Attributes of Agent
Agent Behaviors and Rules
‍‍
Behavior 1: Looking For The Mesh Agents Of The Closest Color
Behavior 2: Moving Towards The Target Mesh Face Agent
Behavior 3: Color-Influencing The Mesh Agents That Are Being Passed By
Behavior 4: Being Deactivated
Similarity-based Agent Model:
System
The system halts when all mesh faces surpass their "active threshold" values, at which point no further agents are activated, and the system deactivates.

‍
Simulation

‍

  • agent num = 5
  • Active Threshold = 2
  • Decay Rate = 80%

‍

  • agent num = 10
  • Active Threshold = 2
  • Decay Rate = 80%
Result