On August 10, 2026, Tran Van Bang successfully defended his Master’s thesis entitled “Text-Based Person Search with Limited Image–Text Pairs.”

Existing Text-Based Person Search (TBPS) methods typically require large-scale training datasets, where each person image is associated with one or more corresponding textual descriptions. However, constructing such image–text pairs requires considerable time and manual annotation effort, especially for large-scale datasets.

To address this limitation, the thesis investigated and developed a person search approach for scenarios where only a limited number of annotated image–text pairs are available. The proposed approach leverages automatic image captioning mechanisms to generate textual descriptions for person images, thereby enriching the training data, reducing the dependence on manual annotation, and improving the model’s ability to learn under limited labeled data.

The thesis contributes to the research direction of Text-Based Person Search under limited data, a practically relevant problem for developing person search and retrieval systems based on natural-language descriptions.

Congratulations to Tran Van Bang on successfully completing this important milestone in his academic and research journey!