安全检查违禁品箱包智能识别数据集

对同一类型的违禁品图片数据进行预处理，按照统一特征值进行归类和提取，最终筛选其能够识别的典型特征归类为一个数据集。
采用端到端基于回归的目标检测和识别，将物体检测作为回归问题求解，将候选框和对象识别两个阶段整合为一个阶段，直接输入图像进行推理，同时得到图像中所有物体的位置和其所属类别及相应的置信概率。

Preprocess the image data of the same type of contraband, classify and extract them based on unified feature values, and finally screen out their recognizable typical features to form a dedicated dataset.
This dataset applies an end-to-end regression-based object detection and recognition framework, which formulates object detection as a regression problem, integrates the two stages of candidate box generation and object recognition into a single stage, directly takes images as input for inference, and simultaneously outputs the positions of all objects in the image, their respective categories, and corresponding confidence probabilities.

搜集汇总

数据集介绍

特点

该数据集包含60000条违禁品数据，涵盖多种违禁品类型，适用于箱包安检设备的智能识别训练，能有效提高安检效率和降低运营成本。