The core difficulty of text-based person search is how to achieve fine-grained alignment of visual and linguistic modal data, so as to bridge the gap of modal heterogeneity. Most existing works on this task focus on global and local features extraction and matching, ignoring the importance of ...
The current text-based person re-identification (re-ID) models tend to learn salient features of image and text, which however is prone to failure in identifying persons with very similar dress, because their image contents with observable but indescribable difference may have identical textual ...
本文认为GNA-RNN方法存在一些不足:1)对图像空间中的关键属性感知能力不强(如文本是“黄色衬衫”,会将“黄色短裤”的行人误判);2)对image-word pair关联度过于敏感(对于一个关键词分配的权重可能会非常大。当一张图像完美匹配了绝大部分关键词时,可能会比匹配所有关键词但不完美时关联度更高)。 本文提出了一个...
Introduction 1) Motivation: 现有大部分visual textual跨模态方法只采用了单一尺度的特征,比如只采用全局尺度或者只采用局部尺度。本文提出了一种动态对齐图文多尺度特征的方法:Non-local Alignment over Full-Scale Represent
RaSa: Relation and Sensitivity Aware Representation Learning for Text-based Person Search RaSa, IJCAI2023 Introduction 基于文本的行人检索具有较大类内方差,和较小类间方差,因此本文提出Relation-aware learning(RA) 和 Sensitivity-aware learning(SA) 着重建模其细粒度关系和提取鉴别性特征。
Text-based person search is a critical task in intelligent security, designed to locate a person of interest by text descriptions. The primary challenge in this task is to effectively bridge the significant gap between the text and image domains while simultaneously extracting the discriminative featur...
This is the official PyTorch implementation of the paperRaSa: Relation and Sensitivity Aware Representation Learning for Text-based Person Search (IJCAI 2023). This repository supports training and evaluation on three text-based person search benchmarks: CUHK-PEDES, ICFG-PEDES and RSTPReid. ...
Text-based person search, employing free-form text queries to identify individuals within a vast image collection, presents a unique challenge in aligning visual and textual representations, particularly at the human part level. Existing methodsoften struggle with part feature extraction and alignment due...
Dual-path CNN with Max Gated block for Text-Based Person Re-identification 作者Tinghuai Ma,∗ , Mingming Yang , Huan Rong , Yurong Qian , Yuan Tiand , NajlaAl-Nabhan 论文链接 arxiv.org/pdf/2009.0934 代码github.com/voriarty/Dua 这篇论文,实验进行了半年的时间,在暑假刚刚完成的,目前是期刊...
Text-Based Person Search with Limited Data brandonhanx/textreid • • 20 Oct 2021 Firstly, to fully utilize the existing small-scale benchmarking datasets for more discriminative feature learning, we introduce a cross-modal momentum contrastive learning framework to enrich the training data for ...