HouseCat6D
收藏资源简介:
HouseCat6D是一个大规模多模态类别级6D物体感知数据集,包含194个不同光度复杂性的家庭对象类别和大量场景,覆盖广泛的视角分布。数据集由慕尼黑工业大学创建,包含同步的RGB、深度和极化RGB+P图像,适用于无纹理、强反射或半透明物体的场景。创建过程中,使用外部红外跟踪系统和稀疏束调整进行后续处理,以避免由时间戳偏移和运动模糊引起的误差。HouseCat6D的应用领域包括日常家庭环境中的类别级姿态估计,旨在解决现有数据集在规模、准确性和现实性方面的局限性。
HouseCat6D is a large-scale multimodal category-level 6D object perception dataset. It encompasses 194 household object categories with diverse photometric complexities, along with numerous scenes covering a broad range of viewpoint distributions. Developed by the Technical University of Munich, this dataset provides synchronized RGB, depth, and polarized RGB+P images, and is applicable to scenarios involving textureless, highly reflective or translucent objects. During the dataset construction, an external infrared tracking system and sparse bundle adjustment were adopted for post-processing to eliminate errors caused by timestamp offsets and motion blur. The application scope of HouseCat6D covers category-level pose estimation in daily household environments, and it is designed to address the limitations of existing datasets in terms of scale, accuracy and realism.




