Distilling Vision-language Pretraining For Efficient Cross-modal Retrieval

Abstract

``Learning to hash'' is a practical solution for efficient retrieval, offering fast search speed and low storage cost. It is widely applied in various applications, such as image-text cross-modal search. In this paper, we explore the potential of enhancing the performance of learning to hash with the proliferation of powerful large pre-trained models, such as Vision-Language Pre-training (VLP) models. We introduce a novel method named Distillation for Cross-Modal Quantization (DCMQ), which leverages the rich semantic knowledge of VLP models to improve hash representation learning. Specifically, we use the VLP as a `teacher' to distill knowledge into a `student' hashing model equipped with codebooks. This process involves the replacement of supervised labels, which are composed of multi-hot vectors and lack semantics, with the rich semantics of VLP. In the end, we apply a transformation termed Normalization with Paired Consistency (NPC) to achieve a discriminative target for distillatio

Stats

• citations0
• S2 citations—
• github stars0
• HF likes0
• heat score0.00
• arxiv keyjang2024distilling

Related papers

• Lightweight Contrastive Distilled Hashing For Online Cross-modal Retrieval (2025)4.52
• Dynamic Contrastive Distillation For Image-text Retrieval (2022)11.76
• Sparse And Dense Retrievers Learn Better Together: Joint Sparse-dense Optimization For Text-image Retrieval (2025)0.00
• Discriminative Cross-view Binary Representation Learning (2018)4.52
• Learning Discriminative Hashing Codes For Cross-modal Retrieval Based On Multi-view Features (2018)3.58
• Deep Collaborative Discrete Hashing With Semantic-invariant Structure (2019)8.09
• Prompthash: Affinity-prompted Collaborative Cross-modal Learning For Adaptive Hashing Retrieval (2025)7.70
• Unsupervised Multi-modal Hashing For Cross-modal Retrieval (2019)8.35