What question did this study set out to answer?

The aim is to improve text-based person search by fully utilizing textual potential from vision-language pre-training.

March 8, 2023Open Access

Exploiting the Textual Potential from Vision-Language Pre-training for Text-based Person Search

View Full Paper

Ask AI

Bookmark

Authors

Randomized trial explores enhanced outcomes in text-based person search, indicating significant performance improvements.

The aim is to improve text-based person search by fully utilizing textual potential from vision-language pre-training.
Developed a baseline model utilizing both visual and textual encoders from vision-language pre-training.
Introduced Multi-Integrity Description Constraints (MIDC) to improve robustness during training.
Implemented Dynamic Attribute Prompt (DAP) for providing fine-grained language hints to the image modality.
Achieved state-of-the-art performance surpassing the previous best method.
Demonstrated effective utilization of textual representations alongside visual encoders.

Wang et al. (2023) studied this question.

View Full Paper

Ask AI

Bookmark