my-data

Name: my-data
Creator: maas
Published: 2025-05-30 20:36:02
License: 暂无描述

魔搭社区2025-05-30 更新2025-05-17 收录

下载链接：

https://modelscope.cn/datasets/leowizard/my-data

下载链接

链接失效反馈

官方服务：

资源简介：

### Dataset Card for STCray   Advancements in Computer-Aided Screening (CAS) systems are essential for improving the detection of security threats in X-ray baggage scans. However, current datasets are limited in representing real-world, sophisticated threats and concealment tactics, and existing approaches are constrained by a closed-set paradigm with predefined labels. To address these challenges, we introduce STCray, the first multimodal X-ray baggage security dataset, comprising 46,642 image-caption paired scans across 21 threat categories, generated using an X-ray scanner for airport security. STCray is meticulously developed with our specialized protocol that ensures domain-aware, coherent captions that lead to the multi-modal instruction following data in X-ray baggage security.         ### Dataset Sources   - **Paper:** The STCray Dataset is part of the paper in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2025 titled "STING-BEE : Towards Vision-Language Model for Real-World X-ray Baggage Security Inspection". https://cvpr.thecvf.com/Conferences/2025/AcceptedPapers - **Github:** https://github.com/Divs1159/STING-BEE - **Website** https://divs1159.github.io/STING-BEE/         ## STCray Dataset Characteristics The Strategic Threat Concealment X-ray (STCray) dataset introduces unique challenges that set it apart from existing X-ray security benchmarks, establishing itself as a pivotal resource for advancing research in baggage threat detection. Characteristics and challenges that STCray brings to the research community: **1) Emerging Sophisticated Threats:** STCray addresses modern security challenges by incorporating a diverse range of contemporary threats, including 3D-printed firearms and improvised explosive devices (IEDs), along side other prohibited items. Specifically, the dataset features three distinct 3D-printed gun designs: the single shot Liberator, the minimalist and recent Harlot, and the pepper-box styled Maverick. **2) Strategic Threat Concealment:** STCray is meticulously designed to reflect the complexities encountered in real-world baggage screening scenarios, using a carefully designed STING protocol, incorporating systematic threat concealment strategies that mimic smuggling tactics and concealment practices.  **3) Diversity and Realism:** Unlike traditional datasets that often retain similar configurations across train and test sets—varying primarily in occlusion levels but preserving the same benign objects and threat items—STCray was designed to reflect the complexity and variability of real-world baggage scenarios. ## Xray Scanner Details The STCray dataset is collected using an ANER K8065 X-ray scanner. It is an advanced X-ray baggage scanner designed for security applications in airports, metro stations, and other security checkpoints. ## Dataset Structure The STCray dataset features a wide spectrum of threat categories, ranging from sharp metallic threats to flammable materials, including Explosive, Gun, 3D-printed Gun, Knife, Cutter, Blade, Shaving Razor, Lighter, Syringe, Battery, Nail Cutter, Other Sharp Item, Powerbank, Scissors, Hammer, Pliers, Wrench, Screwdriver, Handcuffs, and Bullet.   # Dataset Split Our STCray dataset is comprehensively annotated, featuring 46,642 samples with detailed descriptions, bounding boxes, and pixel-level labels. The dataset comprises 30,044 training and 16,598 test images, ensuring a robust evaluation set. Each dataset image is paired with ground truth for classification, bounding box, segmentation mask and detailed textural caption.   ## Citation **BibTeX:** ```bibtex {@Inproceedings{Velayudhan2025StingBee, author = {Divya Velayudhan and Abdelfatah Ahmed and Mohamad Alansari and Neha Gour and Abderaouf Behouch and Taimur Hassan and Syed Talal Wasim and Nabil Maalej and Muzammal Naseer and Jürgen Gall and Mohammed Bennamoun and Ernesto Damiani and Naoufel Werghi}, title = {STING-BEE : Towards Vision-Language Model for Real-World X-ray Baggage Security Inspection}, booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)}, year = {2025}, }

# STCray 数据集卡片   计算机辅助安检（Computer-Aided Screening, CAS）系统的迭代升级，对于提升X射线行李安检扫描中的安全威胁检测能力至关重要。然而，现有数据集在还原真实场景下的复杂威胁与藏匿手段方面存在明显局限，且现有研究方法多受限于带有预定义标签的闭集范式。为解决上述痛点，我们推出了**STCray**——首个多模态X射线行李安检数据集。该数据集依托机场安检专用X射线扫描仪生成，涵盖21类威胁场景，总计46642组图文配对扫描样本。STCray通过定制化开发流程严格构建，确保生成的字幕具备领域适配性与语义连贯性，可支撑X射线行李安检场景下的多模态指令跟随数据研究。         ## 数据集来源   - **论文：** STCray 数据集收录于2025年IEEE/CVF计算机视觉与模式识别会议（CVPR 2025）论文集，论文标题为《STING-BEE : Towards Vision-Language Model for Real-World X-ray Baggage Security Inspection》，会议官网链接：https://cvpr.thecvf.com/Conferences/2025/AcceptedPapers - **GitHub仓库：** https://github.com/Divs1159/STING-BEE - **项目官网：** https://divs1159.github.io/STING-BEE/          ## STCray 数据集特性战略威胁藏匿X射线（Strategic Threat Concealment X-ray, STCray）数据集带来了独有的挑战，使其区别于现有X射线安检基准数据集，成为推动行李威胁检测研究的关键资源。STCray为科研社区带来的特性与挑战如下： **1) 新兴复杂威胁**：STCray纳入了多样化的当代安全威胁，包括3D打印枪械（3D-printed Firearm）、简易爆炸装置（Improvised Explosive Device, IED）及其他违禁物品，以应对现代安检难题。具体而言，数据集涵盖三款独特的3D打印枪械设计：单发式Liberator手枪、极简新型Harlot枪械，以及胡椒盒式Maverick枪械。 **2) 战略性威胁藏匿**：STCray通过精心设计的STING协议，系统模拟走私藏匿策略与实际藏匿手段，精准还原真实行李安检场景中的复杂态势，其开发流程经过严格打磨。 **3) 多样性与真实性**：传统数据集的训练集与测试集往往配置相似，仅在遮挡程度上存在差异，且保留相同的常规无害物品与威胁物品。与之不同，STCray的设计充分贴合真实行李安检场景的复杂性与多变性。  ## X射线扫描仪详情 STCray数据集采用ANER K8065型X射线扫描仪采集，该设备为面向机场、地铁站及其他安检卡口的高端X射线行李安检扫描仪。 ## 数据集结构 STCray数据集覆盖全品类威胁类别，涵盖锋利金属威胁、易燃易爆材料等，具体包括：爆炸物、枪械、3D打印枪械、刀具、切割工具、刀片、剃须刀、打火机、注射器、电池、指甲剪、其他锋利物品、移动电源、剪刀、锤子、钳子、扳手、螺丝刀、手铐以及子弹。   ## 数据集划分 STCray数据集经过全面标注，总计46642个样本，均附带详细描述、边界框、像素级标签。数据集划分为30044张训练图像与16598张测试图像，以保障评估集的可靠性。每张数据集图像均配有分类、边界框、分割掩码以及详细文本描述的真实标注。  ## 引用 **BibTeX格式引用：** bibtex @Inproceedings{Velayudhan2025StingBee, author = {Divya Velayudhan and Abdelfatah Ahmed and Mohamad Alansari and Neha Gour and Abderaouf Behouch and Taimur Hassan and Syed Talal Wasim and Nabil Maalej and Muzammal Naseer and Jürgen Gall and Mohammed Bennamoun and Ernesto Damiani and Naoufel Werghi}, title = {STING-BEE : Towards Vision-Language Model for Real-World X-ray Baggage Security Inspection}, booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)}, year = {2025}, }

提供机构：

maas

创建时间：

2025-04-25

5,000+

优质数据集

54 个

任务类型

进入经典数据集