T-LESS Textureless Object Detection and 6D Pose Estimation Dataset

T-LESS is an RGB-D dataset used for textureless object detection and 6D pose estimation, which is used to estimate the 6D pose of textureless rigid objects. This dataset has 30 objects from different industries. Since there is no obvious texture, discernible color and reflection characteristics, the objects show symmetry and similarity in shape and size.

Compared with other datasets, a major feature of T-LESS is that there is a high degree of overlap between objects. In addition, the dataset also includes training images captured by three synchronized sensors. The data sources include structured light, RGB-D sensor and high-resolution camera. Each sensor has approximately 39k training images and 10k test images.

The T-LESS dataset was released by the Center for Machine Perception of the Czech Technical University and the Institute of Computer Science of the Czech Technical University in 2017. The project leaders are T. Hodaň, P. Haluza, Š. Obdržálek, J. Matas, M. Lourakis and X. Zabulis.

The initial version of this dataset comes from the DARWIN project, and the related paper is "T-LESS: An RGB-D Dataset for 6D Pose Estimation of Texture-less Objects".