SYNTHIA: Novel Concept Design with Affordance Composition

Abstract: Text-to-image (T2I) models enable rapid concept design, making them widely used in AI-driven design. While recent studies focus on generating semantic and stylistic variations of given design concepts, functional coherence--the integration of multiple affordances into a single coherent concept--remains largely overlooked. In this paper, we introduce SYNTHIA, a framework for generating novel, functionally coherent designs based on desired affordances. Our approach leverages a hierarchical concept ontology that decomposes concepts into parts and affordances, serving as a crucial building block for functionally coherent design. We also develop a curriculum learning scheme based on our ontology that contrastively fine-tunes T2I models to progressively learn affordance composition while maintaining visual novelty. To elaborate, we (i) gradually increase affordance distance, guiding models from basic concept-affordance association to complex affordance compositions that integrate parts of distinct affordances into a single, coherent form, and (ii) enforce visual novelty by employing contrastive objectives to push learned representations away from existing concepts. Experimental results show that SYNTHIA outperforms state-of-the-art T2I models, demonstrating absolute gains of 25.1% and 14.7% for novelty and functional coherence in human evaluation, respectively.

Synthia: Novel Concept Design with Affordance Composition

In the field of AI-driven design, Text-to-Image (T2I) models have significantly advanced the ability to generate novel visual concepts. While existing methodologies emphasize semantic and stylistic variations, the capability of these models to integrate multiple affordances—thereby ensuring functional coherence—has been largely understudied. This paper presents "Synthia," a framework that addresses this gap by facilitating the generation of functionally coherent and visually novel designs based on specified affordances. The approach is predicated on a hierarchical concept ontology that associates concepts with their constituent parts and corresponding affordances, serving as an integral framework for synthesizing novel designs with functional coherence.

Synthia introduces an innovative curriculum learning scheme that utilizes the constructed ontology to fine-tune T2I models. The learning involves progressively increasing affordance distance—guiding models from elementary concept-affordance associations to intricate affordance compositions. This methodology supports integrating parts from distinct affordances into a single coherent design, while simultaneously maintaining visual novelty through contrastive objectives. The model effectively ensures that learned representations are distinct from existing concepts by diverting them away from such associations, ultimately fostering the development of novel designs.

Experimental results underscore the efficacy of Synthia in comparison to state-of-the-art T2I models. The framework has demonstrated substantial improvements in human evaluations, showing absolute gains of 25.1% and 14.7% in terms of novelty and functional coherence, respectively. These findings illustrate Synthia's superior capability in generating novel concepts that not only adhere closely to the predefined affordances but also enhance practical applicability and coherence.

The implications of this research extend into both practical and theoretical domains. Practically, Synthia paves the way for advanced design applications that require the blending of unique functionalities into coherent novel forms—ranging from industrial design to creative arts. Theoretically, it contributes to a deeper understanding of affordance integration and its role in achieving functional coherence in AI-generated designs.

Looking towards the future, developments in AI could leverage the framework provided by Synthia to explore more diverse and complex affordance compositions, potentially expanding to broader domains and more intricate functionalities. Further research may also explore the integration of real-world user feedback into the design process, thus refining the quality and applicability of AI-generated designs. Synthia stands as a promising advancement in the intersection of AI design and functional coherence, marking a significant stride in concept synthesis guided by structured affordance-driven models.