Language-Agnostic Visual Embeddings for Cross-Script Handwriting Retrieval
This work addresses the challenge of resource-efficient cross-script handwriting retrieval for digital archives, representing an incremental improvement over existing methods.
The paper tackles the problem of cross-script handwritten word retrieval by proposing a lightweight asymmetric dual-encoder framework that learns unified, style-invariant visual embeddings, achieving state-of-the-art accuracy on within-language benchmarks and strong performance in cross-lingual retrieval with reduced computational costs.
Handwritten word retrieval is vital for digital archives but remains challenging due to large handwriting variability and cross-lingual semantic gaps. While large vision-language models offer potential solutions, their prohibitive computational costs hinder practical edge deployment. To address this, we propose a lightweight asymmetric dual-encoder framework that learns unified, style-invariant visual embeddings. By jointly optimizing instance-level alignment and class-level semantic consistency, our approach anchors visual embeddings to language-agnostic semantic prototypes, enforcing invariance across scripts and writing styles. Experiments show that our method outperforms 28 baselines and achieves state-of-the-art accuracy on within-language retrieval benchmarks. We further conduct explicit cross-lingual retrieval, where the query language differs from the target language, to validate the effectiveness of the learned cross-lingual representations. Achieving strong performance with only a fraction of the parameters required by existing models, our framework enables accurate and resource-efficient cross-script handwriting retrieval.