Highreflection and texture-less workpiece dataset(HW6D) for 6D pose estimation

This is a selfmade datases used to evaluate the 6D pose estimation network named Swin-AFF for highreflection and texture-less workpieces. In this dataset, flange and connector are metal workpieces with reflective surfaces and symmetrical shapes, while frame and clamp are 3D printed plastic workpieces without texture, varying in sizes and shapes. The Intel RealSense D435i camera was used to collect 8 videos. The actual pose of the target workpiece in each frame was labeled, which produced approximately 6,000 labeled images for each workpiece. The dataset was divided according to LineMOD, and an additional 20,000 synthetic data were added for each object to expand the dataset本数据集为自制数据集，用于评估面向高反射无纹理工件的6D位姿估计网络Swin-AFF。其中，法兰盘（flange）与连接器（connector）为表面具有反射特性且形状对称的金属工件，支架（frame）与夹具（clamp）为无纹理、尺寸与形状各异的3D打印塑料工件。本数据集采用英特尔实感D435i（Intel RealSense D435i）相机采集了8段视频，对每帧图像中的目标工件实际位姿进行标注，最终为每个工件生成约6000张标注图像。数据集按照LineMOD的划分规则进行拆分，并为每个目标物体额外添加20000条合成数据以扩充数据集。