{"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/hd-cnn-hierarchical-deep-convolutional-neural","title":"HD-CNN: Hierarchical Deep Convolutional Neural Network for Large Scale Visual Recognition","arxiv_id":"1410.0736","date":"2014-10-03","proceeding":null,"authors":["Zhicheng Yan","Hao Zhang","Robinson Piramuthu","Vignesh Jagadeesh","Dennis Decoste","Wei Di","Yizhou Yu"],"abstract":"In image classification, visual separability between different object\ncategories is highly uneven, and some categories are more difficult to\ndistinguish than others. Such difficult categories demand more dedicated\nclassifiers. However, existing deep convolutional neural networks (CNN) are\ntrained as flat N-way classifiers, and few efforts have been made to leverage\nthe hierarchical structure of categories. In this paper, we introduce\nhierarchical deep CNNs (HD-CNNs) by embedding deep CNNs into a category\nhierarchy. An HD-CNN separates easy classes using a coarse category classifier\nwhile distinguishing difficult classes using fine category classifiers. During\nHD-CNN training, component-wise pretraining is followed by global finetuning\nwith a multinomial logistic loss regularized by a coarse category consistency\nterm. In addition, conditional executions of fine category classifiers and\nlayer parameter compression make HD-CNNs scalable for large-scale visual\nrecognition. We achieve state-of-the-art results on both CIFAR100 and\nlarge-scale ImageNet 1000-class benchmark datasets. In our experiments, we\nbuild up three different HD-CNNs and they lower the top-1 error of the standard\nCNNs by 2.65%, 3.1% and 1.1%, respectively.","url_abs":"http://arxiv.org/abs/1410.0736v4","url_pdf":"http://arxiv.org/pdf/1410.0736v4.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":"hd-cnn-hierarchical-deep-convolutional-neural","repo_url":"https://github.com/Jaehoon-Cha-Data/HD-CNN-Hierarchical-Deep-Convolutional-Neural-Network","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null},{"paper_slug":"hd-cnn-hierarchical-deep-convolutional-neural","repo_url":"https://github.com/justinessert/hierarchical-deep-cnn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"hd-cnn-hierarchical-deep-convolutional-neural","repo_url":"https://github.com/rarriaza/ATPRO_HCNN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null},{"paper_slug":"hd-cnn-hierarchical-deep-convolutional-neural","repo_url":"https://github.com/simpleintel/jupyter","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"Apache-2.0"}}],"tasks":[{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"object-recognition","task_name":"Object Recognition"},{"task_slug":"image-classification","task_name":"image-classification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/image-classification-on-cifar-100","task":"Image Classification","dataset":"CIFAR-100","model":"HD-CNN","rank_in_archive_order":185,"of":211,"metrics":{"Percentage correct":"67.4"},"uses_additional_data":false}],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}