A method for text-based person search in Vietnamese language based on correlation filtering

Text-based person search that aims to associate pedestrian images with free-form natural language descriptions has recently emerged thanks to its wide range of applications, such as searching for missing people and tracking criminals. While the majority of existing methods focus on English and obtain promising results, text-based person in other languages, including Vietnamese, is still in its infancy due to the different characteristics of the languages. In this paper, we propose a method for person search through queries in Vietnamese. In this method, to tackle the specific characteristics of Vietnamese, a Vietnamese language parser and textual feature extractor have been integrated into a model based on the correlation filtering named SRCF that has been proposed for text-based person search in English [1] which named Correlation Filter For Vietnamese Language (CFFVL). Extensive experiments have been conducted on VNPersonSearch3000, a large-scale dataset for person search in Vietnam, to evaluate the effectiveness of different Vietnamese language parsers and textual feature extractors. The experimental results show that the proposed method outperforms the state-of-the-art method by 18.42% in top-1 and 23.10% in top-5 on the VNPersonSearch3000 dataset.