Cross-modal image-text retrieval is a fundamental task in bridging vision and language. It faces two main challenges that are typically not well addressed in previous works. 1) Generalizability: Existing methods often...Cross-modal image-text retrieval is a fundamental task in bridging vision and language. It faces two main challenges that are typically not well addressed in previous works. 1) Generalizability: Existing methods often assume a strong semantic correlation between each text-image pair, which are thus difficult to generalize to real-world scenarios where the weak correlation dominates. 2) Efficiency: Many latest works adopt the single-tower architecture with heavy detectors, which are inefficient during the inference stage because the costly computation needs to be repeated for each text-image pair. In this work, to overcome these two challenges, we propose a two-tower cross-modal contrastive learning (CMCL) framework. Specifically, we first devise a two-tower architecture, which enables a unified feature space for the text and image modalities to be directly compared with each other, alleviating the heavy computation during inference. We further introduce a simple yet effective module named multi-grid split (MGS) to learn fine-grained image features without using detectors. Last but not the least, we deploy a cross-modal contrastive loss on the global image/text features to learn their weak correlation and thus achieve high generalizability. To validate that our CMCL can be readily generalized to real-world scenarios, we construct a large multi-source image-text dataset called weak semantic correlation dataset (WSCD). Extensive experiments show that our CMCL outperforms the state-of-the-arts while being much more efficient.展开更多
To the Editor:Drug resistance mutations(DRMs)involving human deficiency virus type 1(HIV-1)diminish the efficacy of current antiretroviral therapy(ART)regimens,resulting in treatment failure.A better understanding of ...To the Editor:Drug resistance mutations(DRMs)involving human deficiency virus type 1(HIV-1)diminish the efficacy of current antiretroviral therapy(ART)regimens,resulting in treatment failure.A better understanding of the prevalence of local DRMs that lead to ART failure can provide valuable data for clinical-and government-level decision-making.The DRM status of people living with HIV(PLWH)refractory to ART(viral load>200 copies/mL)was assessed based on analysis of plasma samples derived from PLWH in the Sixth People’s Hospital of Zhengzhou from June 2018 to April 2022.Further,the partial polymerase(pol)and complete integrase gene-coding sequences were amplified,sequenced,and analyzed to identify DRMs.This study was approved by the Institutional Ethics Committee of the Sixth People’s Hospital of Zhengzhou(No.2019-04).All participants signed written informed consent before sample collection.展开更多
To the Editor:Even with the emergence of highly active antiretroviral therapy,human immunodeficiency virus type 1(HIV-1)remains a serious public health problem worldwide.The widespread use of antiretroviral therapy(AR...To the Editor:Even with the emergence of highly active antiretroviral therapy,human immunodeficiency virus type 1(HIV-1)remains a serious public health problem worldwide.The widespread use of antiretroviral therapy(ART)has,on the one hand,significantly improved life expectancy in patients infected with HIV-1;on the other hand,it has inevitably brought along with it a difficult problem—viral resistance.Viral resistance is a common phenomenon evolutionarily,particularly in patients with poor adherence and suboptimal drug levels.展开更多
文摘Cross-modal image-text retrieval is a fundamental task in bridging vision and language. It faces two main challenges that are typically not well addressed in previous works. 1) Generalizability: Existing methods often assume a strong semantic correlation between each text-image pair, which are thus difficult to generalize to real-world scenarios where the weak correlation dominates. 2) Efficiency: Many latest works adopt the single-tower architecture with heavy detectors, which are inefficient during the inference stage because the costly computation needs to be repeated for each text-image pair. In this work, to overcome these two challenges, we propose a two-tower cross-modal contrastive learning (CMCL) framework. Specifically, we first devise a two-tower architecture, which enables a unified feature space for the text and image modalities to be directly compared with each other, alleviating the heavy computation during inference. We further introduce a simple yet effective module named multi-grid split (MGS) to learn fine-grained image features without using detectors. Last but not the least, we deploy a cross-modal contrastive loss on the global image/text features to learn their weak correlation and thus achieve high generalizability. To validate that our CMCL can be readily generalized to real-world scenarios, we construct a large multi-source image-text dataset called weak semantic correlation dataset (WSCD). Extensive experiments show that our CMCL outperforms the state-of-the-arts while being much more efficient.
基金supported by grants from the Science and Technology Research Project of Henan Province(Nos.SB201903031 and 232102310203)
文摘To the Editor:Drug resistance mutations(DRMs)involving human deficiency virus type 1(HIV-1)diminish the efficacy of current antiretroviral therapy(ART)regimens,resulting in treatment failure.A better understanding of the prevalence of local DRMs that lead to ART failure can provide valuable data for clinical-and government-level decision-making.The DRM status of people living with HIV(PLWH)refractory to ART(viral load>200 copies/mL)was assessed based on analysis of plasma samples derived from PLWH in the Sixth People’s Hospital of Zhengzhou from June 2018 to April 2022.Further,the partial polymerase(pol)and complete integrase gene-coding sequences were amplified,sequenced,and analyzed to identify DRMs.This study was approved by the Institutional Ethics Committee of the Sixth People’s Hospital of Zhengzhou(No.2019-04).All participants signed written informed consent before sample collection.
基金Co-Sponsored Construction Project of Provincial Medicine and Technique of Henan Province(No. LHGJ20191103)。
文摘To the Editor:Even with the emergence of highly active antiretroviral therapy,human immunodeficiency virus type 1(HIV-1)remains a serious public health problem worldwide.The widespread use of antiretroviral therapy(ART)has,on the one hand,significantly improved life expectancy in patients infected with HIV-1;on the other hand,it has inevitably brought along with it a difficult problem—viral resistance.Viral resistance is a common phenomenon evolutionarily,particularly in patients with poor adherence and suboptimal drug levels.