PhoCaL

PhoCaL数据集由慕尼黑工业大学创建，包含60个高质量的3D模型，涵盖8个类别，包括反射性、透明和对称等具有光度学挑战性的物体。数据集通过创新的机器人辅助多模态（RGB、深度、偏振）数据采集和标注过程，确保了6D姿态标注的亚毫米级精度。该数据集适用于机器人应用和增强现实中的类别级物体姿态估计，旨在解决光度学挑战性物体在实际操作环境中的姿态估计问题。

The PhoCaL dataset was developed by the Technical University of Munich. It contains 60 high-quality 3D models spanning 8 categories, including photometrically challenging objects such as reflective, transparent and symmetric ones. Through an innovative robot-assisted multimodal (RGB, depth, polarization) data acquisition and annotation process, the dataset guarantees sub-millimeter precision for 6D pose annotations. This dataset is designed for category-level object pose estimation in robotic applications and augmented reality, aiming to solve the problem of pose estimation for photometrically challenging objects in real-world operational environments.