# A Multihead Continual Learning Framework for Fine-Grained Fashion Image Retrieval with Contrastive Learning and Exponential Moving Average Distillation

arXiv: https://arxiv.org/abs/2603.20648 · HTML: https://mlsanity.com/paper/2603.20648/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 24 of its 48 references that are in the map.
- First posted on arXiv 21 Mar 2026.

## Builds on (most cited first, 24 of 24)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [Experience Replay for Continual Learning](https://mlsanity.com/paper/1811.11682/) (2018)
- [Class-Incremental Learning: A Survey](https://mlsanity.com/paper/2302.03648/) (2023)
- [Expandable Subspace Ensemble for Pre-Trained Model-Based Class-Incremental Learning](https://mlsanity.com/paper/2403.12030/) (2024)
- [Cross-domain Image Retrieval with a Dual Attribute-aware Ranking Network](https://mlsanity.com/paper/1505.07922/) (2015)
- [InfoNCE: Identifying the Gap Between Theory and Practice](https://mlsanity.com/paper/2407.00143/) (2024)
- [A Multi-Head Model for Continual Learning via Out-of-Distribution Replay](https://mlsanity.com/paper/2208.09734/) (2022)
- [Learning Adaptive Embedding Considering Incremental Class](https://mlsanity.com/paper/2008.13351/) (2020)
- [Improving Plasticity in Online Continual Learning via Collaborative Learning](https://mlsanity.com/paper/2312.00600/) (2023)
- [Rethinking Momentum Knowledge Distillation in Online Continual Learning](https://mlsanity.com/paper/2309.02870/) (2023)
- [Neural Compatibility Modeling with Attentive Knowledge Distillation](https://mlsanity.com/paper/1805.00313/) (2018)
- [Fine-Grained Fashion Similarity Learning by Attribute-Specific Embedding Network](https://mlsanity.com/paper/2002.02814/) (2020)
- [FashionSAP: Symbols and Attributes Prompt for Fine-grained Fashion Vision-Language Pre-training](https://mlsanity.com/paper/2304.05051/) (2023)
- [Fine-Grained Fashion Similarity Prediction by Attribute-Specific Embedding Learning](https://mlsanity.com/paper/2104.02429/) (2021)
- [From Region to Patch: Attribute-Aware Foreground-Background Contrastive Learning for Fine-Grained Fashion Retrieval](https://mlsanity.com/paper/2305.10260/) (2023)
- [Learning Diverse Fashion Collocation by Neural Graph Filtering](https://mlsanity.com/paper/2003.04888/) (2020)
- [Query-free Clothing Retrieval via Implicit Relevance Feedback](https://mlsanity.com/paper/1711.00248/) (2017)
- [Training and challenging models for text-guided fashion image retrieval](https://mlsanity.com/paper/2204.11004/) (2022)
- [SAT: Self-adaptive training for fashion compatibility prediction](https://mlsanity.com/paper/2206.12622/) (2022)
- [Attribute-Guided Multi-Level Attention Network for Fine-Grained Fashion Retrieval](https://mlsanity.com/paper/2301.13014/) (2022)
- [Class Balance Matters to Active Class-Incremental Learning](https://mlsanity.com/paper/2412.06642/) (2024)
- [Can GPT-4o mini and Gemini 2.0 Flash Predict Fine-Grained Fashion Product Attributes? A Zero-Shot Analysis](https://mlsanity.com/paper/2507.09950/) (2025)
- [FashionFAE: Fine-grained Attributes Enhanced Fashion Vision-Language Pre-training](https://mlsanity.com/paper/2412.19997/) (2024)

## Built on by (most cited first, 1 of 1)

- [FashionLens: Toward Versatile Fashion Image Retrieval via Task-Adaptive Learning](https://mlsanity.com/paper/2605.22552/) (2026)

## Abstract (arXiv)

Most fine-grained fashion image retrieval (FIR) methods assume a static setting, requiring full retraining when new attributes appear, which is costly and impractical for dynamic scenarios. Although pretrained models support zero-shot inference, their accuracy drops without supervision, and no prior work explores class-incremental learning (CIL) for fine-grained FIR. We propose a multihead continual learning framework for fine-grained fashion image retrieval with contrastive learning and exponential moving average (EMA) distillation (MCL-FIR). MCL-FIR adopts a multi-head design to accommodate evolving classes across increments, reformulates triplet inputs into doublets with InfoNCE for simpler and more effective training, and employs EMA distillation for efficient knowledge transfer. Experiments across four datasets demonstrate that, beyond its scalability, MCL-FIR achieves a strong balance between efficiency and accuracy. It significantly outperforms CIL baselines under similar training cost, and compared with static methods, it delivers comparable performance while using only about 30% of the training cost. The source code is publicly available in https://github.com/Dr-LingXiao/MCL-FIR.