Surgical Waste Dataset

Please cite our paper titled 'A Computer Vision-based System for Surgical Waste Detection'

Digital Object Identifier (DOI) : 10.14569/IJACSA.2022.0130366

The dataset is built based on real-time images from our surroundings including roads, beaches, water, maintenance holes and so on. Several images of the dataset are synthetic. Moreover, most of them are natural. Some images are taken using the Samsung Galaxy A51 smartphone camera and the rest of the images are taken from internet mining. Images are chosen from close range and distance range to make the dataset a distance variant. The angle variation left, right, back and top angle images are taken. The dataset comprises diverse gesture conditions such as curling and kneeling. At the time of image collection, this study tries to take different types of colored masks and gloves. The color variation of the mask is white, sky blue, pink, black and others. Different types of masks are included surgical, N95, Cone-style, KN95 and so on. Surgical gloves also have blue, white, black and pink colors. Transparent gloves are included with more eagerness to make the system as robust and reliable underwater as well as an object floating on the water condition. According to the above criteria, 1153 images are collected from different internet sources and smartphones camera.

请引用我们题为《基于计算机视觉的手术废弃物检测系统》的研究论文。
数字对象标识符（Digital Object Identifier，DOI）：10.14569/IJACSA.2022.0130366

本数据集基于采集自真实场景的实拍图像构建，涵盖道路、海滩、水域、检查井等多种场景；数据集包含部分合成图像，且绝大多数为自然实拍样本。部分图像通过三星Galaxy A51智能手机摄像头拍摄，其余图像则通过网络爬取获取。
为确保数据集具备距离维度的多样性，图像采集覆盖近距与远距范围；同时采集了左、右、背、顶等不同拍摄角度的图像，实现角度维度的变化。数据集涵盖多种拍摄姿态场景，例如蜷缩、跪姿等。
在图像采集阶段，本研究采集了不同色彩与类型的防护口罩及手套样本。口罩色彩涵盖白色、天蓝色、粉色、黑色及其他配色；口罩类型包含医用外科口罩、N95口罩、杯型口罩、KN95口罩等。医用手套则包含蓝色、白色、黑色与粉色等配色，同时额外采集了透明手套样本，以提升系统在水下及水上漂浮物体场景下的鲁棒性与可靠性。
按照上述采集标准，本研究共从网络资源与智能手机摄像头处采集得到1153张图像。