FCBoost-Net: A Generative Network for Synthesizing Multiple Collocated Outfits via Fashion Compatibility Boosting

TL;DR

FCBoost-Net tackles the challenge of generating collocated and diversified outfits from a given item set by coupling per-category GANs with GAN inversion and a recurrent fashion compatibility booster. The method introduces three learning objectives—adversarial realism, diversity via LPIPS, and boosting-based compatibility across rounds—within a unified framework trained on the OutfitSet data. Empirical results show significant gains in visual authenticity (FID), diversity (LPIPS), and compatibility (F$^2$BT) compared to strong multimodal I2I baselines, demonstrating practical value for designers seeking multiple outfit options. The work advances fashion synthesis by enabling one-to-many, set-wise generation with explicit supervision and iterative refinement, potentially accelerating design workflows.

Abstract

Outfit generation is a challenging task in the field of fashion technology, in which the aim is to create a collocated set of fashion items that complement a given set of items. Previous studies in this area have been limited to generating a unique set of fashion items based on a given set of items, without providing additional options to users. This lack of a diverse range of choices necessitates the development of a more versatile framework. However, when the task of generating collocated and diversified outfits is approached with multimodal image-to-image translation methods, it poses a challenging problem in terms of non-aligned image translation, which is hard to address with existing methods. In this research, we present FCBoost-Net, a new framework for outfit generation that leverages the power of pre-trained generative models to produce multiple collocated and diversified outfits. Initially, FCBoost-Net randomly synthesizes multiple sets of fashion items, and the compatibility of the synthesized sets is then improved in several rounds using a novel fashion compatibility booster. This approach was inspired by boosting algorithms and allows the performance to be gradually improved in multiple steps. Empirical evidence indicates that the proposed strategy can improve the fashion compatibility of randomly synthesized fashion items as well as maintain their diversity. Extensive experiments confirm the effectiveness of our proposed framework with respect to visual authenticity, diversity, and fashion compatibility.

Figures (6)

• Figure 1: FCBoost-Net is a novel framework that generates multiple diversified sets of fashion items based on a given set of items. Each synthetic set is collocated and can form a complete outfit alongside the given set. Two examples are given here: (a) synthesizing diverse items of bag and shoe that match the given upper and lower clothing; and (b) synthesizing diverse items of lower clothing and shoe that complement the given upper clothing and bag.
• Figure 2: The overall pipeline of FCBoost-Net. (Here, the fashion items labeled with blue boxes represent the given items, while those enclosed within red boxes indicate the synthesized items.)
• Figure 3: Comparisons between our FCBoost-Net and other multimodal I2I translation methods, which are MUNIT huang2018munit, DRIT DRIT, DRIT++ DRIT_plus, and SAVI2I mao2022continuous, for input based on one, two, or three fashion items. (Zoom in for a better view.)
• Figure 4: Visual comparison of the results for the effectiveness of the diversity loss function: (a) FCBoost-Net without diversity loss, and (b) FCBoost-Net with diversity loss (ours).
• Figure 5: Visual comparison results for the effectiveness of fashion compatibility booster: (a) FCBoost-Net without fashion compatibility booster, and (b) FCBoost-Net with fashion compatibility booster (ours).