Sara Sarto, Marcella Cornia, Lorenzo Baraldi, Alessandro Nicolosi, Rita Cucchiara
{"title":"为图像标题设计检索增强架构","authors":"Sara Sarto, Marcella Cornia, Lorenzo Baraldi, Alessandro Nicolosi, Rita Cucchiara","doi":"10.1145/3663667","DOIUrl":null,"url":null,"abstract":"<p>The objective of image captioning models is to bridge the gap between the visual and linguistic modalities by generating natural language descriptions that accurately reflect the content of input images. In recent years, researchers have leveraged deep learning-based models and made advances in the extraction of visual features and the design of multimodal connections to tackle this task. This work presents a novel approach towards developing image captioning models that utilize an external <i>k</i>NN memory to improve the generation process. Specifically, we propose two model variants that incorporate a knowledge retriever component that is based on visual similarities, a differentiable encoder to represent input images, and a <i>k</i>NN-augmented language model to predict tokens based on contextual cues and text retrieved from the external memory. We experimentally validate our approach on COCO and nocaps datasets and demonstrate that incorporating an explicit external memory can significantly enhance the quality of captions, especially with a larger retrieval corpus. This work provides valuable insights into retrieval-augmented captioning models and opens up new avenues for improving image captioning at a larger scale.</p>","PeriodicalId":50937,"journal":{"name":"ACM Transactions on Multimedia Computing Communications and Applications","volume":"11 1","pages":""},"PeriodicalIF":5.2000,"publicationDate":"2024-05-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Towards Retrieval-Augmented Architectures for Image Captioning\",\"authors\":\"Sara Sarto, Marcella Cornia, Lorenzo Baraldi, Alessandro Nicolosi, Rita Cucchiara\",\"doi\":\"10.1145/3663667\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p>The objective of image captioning models is to bridge the gap between the visual and linguistic modalities by generating natural language descriptions that accurately reflect the content of input images. In recent years, researchers have leveraged deep learning-based models and made advances in the extraction of visual features and the design of multimodal connections to tackle this task. This work presents a novel approach towards developing image captioning models that utilize an external <i>k</i>NN memory to improve the generation process. Specifically, we propose two model variants that incorporate a knowledge retriever component that is based on visual similarities, a differentiable encoder to represent input images, and a <i>k</i>NN-augmented language model to predict tokens based on contextual cues and text retrieved from the external memory. We experimentally validate our approach on COCO and nocaps datasets and demonstrate that incorporating an explicit external memory can significantly enhance the quality of captions, especially with a larger retrieval corpus. This work provides valuable insights into retrieval-augmented captioning models and opens up new avenues for improving image captioning at a larger scale.</p>\",\"PeriodicalId\":50937,\"journal\":{\"name\":\"ACM Transactions on Multimedia Computing Communications and Applications\",\"volume\":\"11 1\",\"pages\":\"\"},\"PeriodicalIF\":5.2000,\"publicationDate\":\"2024-05-03\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"ACM Transactions on Multimedia Computing Communications and Applications\",\"FirstCategoryId\":\"94\",\"ListUrlMain\":\"https://doi.org/10.1145/3663667\",\"RegionNum\":3,\"RegionCategory\":\"计算机科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"COMPUTER SCIENCE, INFORMATION SYSTEMS\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"ACM Transactions on Multimedia Computing Communications and Applications","FirstCategoryId":"94","ListUrlMain":"https://doi.org/10.1145/3663667","RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"COMPUTER SCIENCE, INFORMATION SYSTEMS","Score":null,"Total":0}
Towards Retrieval-Augmented Architectures for Image Captioning
The objective of image captioning models is to bridge the gap between the visual and linguistic modalities by generating natural language descriptions that accurately reflect the content of input images. In recent years, researchers have leveraged deep learning-based models and made advances in the extraction of visual features and the design of multimodal connections to tackle this task. This work presents a novel approach towards developing image captioning models that utilize an external kNN memory to improve the generation process. Specifically, we propose two model variants that incorporate a knowledge retriever component that is based on visual similarities, a differentiable encoder to represent input images, and a kNN-augmented language model to predict tokens based on contextual cues and text retrieved from the external memory. We experimentally validate our approach on COCO and nocaps datasets and demonstrate that incorporating an explicit external memory can significantly enhance the quality of captions, especially with a larger retrieval corpus. This work provides valuable insights into retrieval-augmented captioning models and opens up new avenues for improving image captioning at a larger scale.
期刊介绍:
The ACM Transactions on Multimedia Computing, Communications, and Applications is the flagship publication of the ACM Special Interest Group in Multimedia (SIGMM). It is soliciting paper submissions on all aspects of multimedia. Papers on single media (for instance, audio, video, animation) and their processing are also welcome.
TOMM is a peer-reviewed, archival journal, available in both print form and digital form. The Journal is published quarterly; with roughly 7 23-page articles in each issue. In addition, all Special Issues are published online-only to ensure a timely publication. The transactions consists primarily of research papers. This is an archival journal and it is intended that the papers will have lasting importance and value over time. In general, papers whose primary focus is on particular multimedia products or the current state of the industry will not be included.