{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/densecap-fully-convolutional-localization","title":"DenseCap: Fully Convolutional Localization Networks for Dense Captioning","arxiv_id":"1511.07571","date":"2015-11-24","proceeding":"CVPR 2016 6","authors":["Justin Johnson","Andrej Karpathy","Li Fei-Fei"],"abstract":"We introduce the dense captioning task, which requires a computer vision\nsystem to both localize and describe salient regions in images in natural\nlanguage. The dense captioning task generalizes object detection when the\ndescriptions consist of a single word, and Image Captioning when one predicted\nregion covers the full image. To address the localization and description task\njointly we propose a Fully Convolutional Localization Network (FCLN)\narchitecture that processes an image with a single, efficient forward pass,\nrequires no external regions proposals, and can be trained end-to-end with a\nsingle round of optimization. The architecture is composed of a Convolutional\nNetwork, a novel dense localization layer, and Recurrent Neural Network\nlanguage model that generates the label sequences. We evaluate our network on\nthe Visual Genome dataset, which comprises 94,000 images and 4,100,000\nregion-grounded captions. We observe both speed and accuracy improvements over\nbaselines based on current state of the art approaches in both generation and\nretrieval settings.","url_abs":"http://arxiv.org/abs/1511.07571v1","url_pdf":"http://arxiv.org/pdf/1511.07571v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"densecap-fully-convolutional-localization","repo_url":"https://github.com/jcjohnson/densecap","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"torch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"dense-captioning","task_name":"Dense Captioning"},{"task_slug":"image-captioning","task_name":"Image Captioning"},{"task_slug":"language-modeling","task_name":"Language Modeling"},{"task_slug":"language-modelling","task_name":"Language Modelling"},{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"retrieval","task_name":"Retrieval"},{"task_slug":"object-detection-1","task_name":"object-detection"}],"methods":[{"method_slug":"speed","method_name":"SPEED"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/dense-captioning-on-visual-genome","task":"Dense Captioning","dataset":"Visual Genome","model":"FCLN","rank_in_archive_order":4,"of":4,"metrics":{"mAP":"5.4"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-visual-genome","task":"Object Detection","dataset":"Visual Genome","model":"AP (%)","rank_in_archive_order":4,"of":4,"metrics":{"MAP":"5.39"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1511.07571","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}