{"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/sixray-a-large-scale-security-inspection-x","title":"SIXray : A Large-scale Security Inspection X-ray Benchmark for Prohibited Item Discovery in Overlapping Images","arxiv_id":"1901.00303","date":"2019-01-02","proceeding":null,"authors":["Caijing Miao","Lingxi Xie","Fang Wan","Chi Su","Hongye Liu","Jianbin Jiao","Qixiang Ye"],"abstract":"In this paper, we present a large-scale dataset and establish a baseline for\nprohibited item discovery in Security Inspection X-ray images. Our dataset,\nnamed SIXray, consists of 1,059,231 X-ray images, in which 6 classes of 8,929\nprohibited items are manually annotated. It raises a brand new challenge of\noverlapping image data, meanwhile shares the same properties with existing\ndatasets, including complex yet meaningless contexts and class imbalance. We\npropose an approach named class-balanced hierarchical refinement (CHR) to deal\nwith these difficulties. CHR assumes that each input image is sampled from a\nmixture distribution, and that deep networks require an iterative process to\ninfer image contents accurately. To accelerate, we insert reversed connections\nto different network backbones, delivering high-level visual cues to assist\nmid-level features. In addition, a class-balanced loss function is designed to\nmaximally alleviate the noise introduced by easy negative samples. We evaluate\nCHR on SIXray with different ratios of positive/negative samples. Compared to\nthe baselines, CHR enjoys a better ability of discriminating objects especially\nusing mid-level features, which offers the possibility of using a\nweakly-supervised approach towards accurate object localization. In particular,\nthe advantage of CHR is more significant in the scenarios with fewer positive\ntraining samples, which demonstrates its potential application in real-world\nsecurity inspection.","url_abs":"http://arxiv.org/abs/1901.00303v1","url_pdf":"http://arxiv.org/pdf/1901.00303v1.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":"sixray-a-large-scale-security-inspection-x","repo_url":"https://github.com/MeioJane/SIXray","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":"object-localization","task_name":"Object Localization"}],"methods":[],"datasets_introduced":[{"slug":"sixray","name":"SIXray","full_name":null}],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1901.00303","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}