id	author	title	date	pages	extension	mime	words	sentence	flesch	summary	cache	txt
fcis-28893	Qiu, Xiangmian; Zhang, Yali; Zhao, Yichen; Li, Jinzhao	A Review of Text-Based Pedestrian Retrieval Methods	2024	10	.pdf	application/pdf	8365	280	40	Methods of Text-based Pedestrian Retrieval Text-based pedestrian retrieval aims to retrieve images of specific pedestrians through textual descriptions, and its core challenge lies in bridging the inherent gap between the two modalities and realizing cross-modal alignment between image features and text features. First bi-directional LSTM extracts text features and understands potential semantic dependencies between words, similar to the visual scene graph module, the text scene graph module detects objects present in the text and predicts the relationships between them, and at the same time, a graph attention convolutional layer is used to obtain structured text semantic features, and finally, the structured semantic text features and the initial text features are fused, which are mapped by the joint embedding layer into the shared feature space.	cache/fcis-28893.pdf	txt/fcis-28893.txt