TrendGen: An Outfit Recommendation and Display System

Theodoros Koukopoulos; Dimos Klimenof; Ioannis Xarchakos

cs.CVMar 28, 2026

TrendGen: An Outfit Recommendation and Display System

Authors: Theodoros Koukopoulos, Dimos Klimenof, Ioannis Xarchakos

Organizations: SabinoDB, Athens, Greece · University of Toronto

Abstract

Recent advances in Computer Vision have significantly improved image understanding and generation, revolutionizing the fashion industry. However, challenges such as inconsistent lighting, non-ideal garment angles, complex backgrounds, and occlusions in raw images hinder their full potential. Overcoming these obstacles is crucial for developing robust fashion AI systems capable of real-world applications. In this paper, we introduce TrendGen, a Fashion AI system designed to enhance online shopping with intelligent outfit recommendations. Deployed on a major e-commerce platform, TrendGen leverages cloth images and product attributes to generate trend-aligned, cohesive outfit suggestions. Additionally, it employs Generative AI to transform raw images into high-quality lay-down views, offering a clear and structured presentation of garments. Our evaluation on production data demonstrates TrendGen's consistent high-quality outfits and lay-down images, marking a significant advancement in AI-driven solutions for fashion retail.

Figures & tables

Panel of Figure 1 — Figure 1. Different garment image views

Panel of Figure 2 — Figure 2. Cloth masking

Figure 4. Attribution Classifier Architecture

Figure 5. Virtual Try-Off Network Architecture

Methodology	Accuracy (%)
FashionCLIP ( Chia et al., 2022 )	73.8
ResNet ( He et al., 2016b ) + Mxbai ( Li and Li, 2023 ) + Classifier	89.1
FashionCLIP + Classifier (Proposed System)	92.3

Table 1. Product Attribution Accuracy

Figure 6. Comparison against TryOffDiff ( Velioglu et al., 2024 ) and MGT ( Velioglu et al., 2025 )

Method	SSIM ↑	FID ↓	KID ↓	LPIPS ↓	DISTS ↓
TileGan ( Zeng et al., 2020 )	70.96	39.8	N/A	N/A	N/A
TryOffDiff ( Velioglu et al., 2024 )	79.5	25.1	8.9	32.4	23.0
MGT ( Velioglu et al., 2025 )	78.1	21.9	8.9	36.3	24.7
Ours	71.99	25.3	2.01	17.22	21.01

Table 2. Virtual Try-Off Performance

Input Type	Accuracy (%)
Ours without STYLERANK	55.7
OutfitTransformer ( Sarkar et al., 2023 )	75.2
Ours (Raw Images + Title + Description)	79.8
Ours (Lay-down Images + Attributes)	91.3

Table 3. Outfit Recommendation Accuracy

Appendix figures & tables5 assets

Supplementary material from the paper’s appendix.

Appendix

Panel of Figure 7 — Figure 7. Three distinct styling approaches, each representing different fashion aesthetics. These are visually connected through color coordination, accessories, and outfit choices.

Figure 8. Qualitative comparison against TryOffDiff ( Velioglu et al., 2024 ) and MGT ( Velioglu et al., 2025 )

Figure 9. Qualitative comparison against TryOffDiff ( Velioglu et al., 2024 ) and MGT ( Velioglu et al., 2025 )

Figure 10. Qualitative comparison against TryOffDiff ( Velioglu et al., 2024 ) and MGT ( Velioglu et al., 2025 )

$\lambda$	0	0.25	0.5	1.0	1.5	2.0	3.0
Accuracy (%)	55.7	78.4	89.7	91.3	87.1	85.9	77.2

Appendix

Table 4. Ablation study on

\lambda

in STYLERANK. Performance is reported as expert-verified outfit approval rate (%).

Multimodal Recommendation

Multimodal recommendation combines user behavior with item content such as images, text, audio, and video to personalize rankings.

See all 48 papers

Tstars-Tryon 1.0: Robust and Realistic Virtual Try-On for Diverse Fashion Items
Apr 21, 2026Mengting Chen, Zhengrui Chen, Yongchao Du +16Virtual Try-On Conditional Image Generation
Alibaba Group
DetailAnywhere: Fashion Detail Generation via Cross-Modal Feature Alignment Distillation
Jul 2, 2026Zijun Li, Yimin Zhou, Jia Sun +12Diffusion Model Distillation Conditional Image Generation
Kuaishou Technology · AIM for Health Lab, Monash University
Exploring Multi-Modal Large Language Models and Two-Stage Fine-Tuning for Fashion Image Retrieval
Jun 18, 2026Nguyen Cao Hoang, Hoang Bui Le, Nam Vo Hoang +1Multimodal Large Language Models Image-Text Retrieval
University of Science, VNU-HCM, Ho Chi Minh, Vietnam · Vietnam National University, Ho Chi Minh, Vietnam

Apr 21, 2026J/K move · Enter open · S save

Tstars-Tryon 1.0: Robust and Realistic Virtual Try-On for Diverse Fashion Items

Mengting Chen, Zhengrui Chen, Yongchao Du, Zuan Gao+15

Alibaba Group

Virtual Try-On Conditional Image Generation Controllable Image Generation

Open paper AskPDF arXiv

Recent advances in image generation and editing have opened new opportunities for virtual try-on. However, existing methods still struggle to meet complex real-world demands. We present Tstars-Tryon 1.0, a commercial-scale virtual try-on system that is robust, realistic, versatile, and highly efficient. First, our system maintains a high success rate across challenging cases like extreme poses, severe illumination variations, motion blur, and other in-the-wild conditions. Second, it delivers highly photorealistic results with fine-grained details, faithfully preserving garment texture, material properties, and structural characteristics, while largely avoiding common AI-generated artifacts. Third, beyond apparel try-on, our model supports flexible multi-image composition (up to 6 reference images) across 8 fashion categories, with coordinated control over person identity and background. Fourth, to overcome the latency bottlenecks of commercial deployment, our system is heavily optimized for inference speed, delivering near real-time generation for a seamless user experience. These capabilities are enabled by an integrated system design spanning end-to-end model architecture, a scalable data engine, robust infrastructure, and a multi-stage training paradigm. Extensive evaluation and large-scale product deployment demonstrate that Tstars-Tryon1.0 achieves leading overall performance. To support future research, we also release a comprehensive benchmark. The model has been deployed at an industrial scale on the Taobao App, serving millions of users with tens of millions of requests.

Abstract

Multimodal Recommendation

Explore similar work

Tstars-Tryon 1.0: Robust and Realistic Virtual Try-On for Diverse Fashion Items

DetailAnywhere: Fashion Detail Generation via Cross-Modal Feature Alignment Distillation

Exploring Multi-Modal Large Language Models and Two-Stage Fine-Tuning for Fashion Image Retrieval