RGB-D dataset for 6D pose estimation of industrial components/tools

• BOP format: This dataset fully complies with the BOP format
• 5 objects commonly found in an industrial assembly environment: The objects are symmetric and textureless presenting challenges for 6D pose estimation.
• Data acquisition with an Intel RealSense D455 camera: The dataset was captured at a resolution of 640×480. Both training and test images were collected under controlled conditions.
• Training images (4,107 in total): 1361 synthetic images, 2.746 real including scenes with individual and multiple objects are depicted
• Test images (1,354 in total): 1354 real test images.

The photorealistic synthetic images were generated using a custom OpenGL-based implementation: 6DL-PoseGenerator [1]

[1] Pateraki, M., Sapoutzoglou, P., Lourakis, M.I.: Crane spreader pose estimation from a single view. In: Proceedings of the 18th International Joint Conference on Computer Vision, vol. 5, pp. 796-805. SCITEPRESS (2023)

6DAPose dataset contains an object assembly in an ordered manner following its assembly steps. Each dataset contains 431 simulated view samples of the assembly for each assembly step. For each view sample, RGB and Depth Image, Segmentation map, ground truth values for 6D object pose and camera pose and model information are recorded following the BOP format . There are two datasets representing a 3-step fidget gear assembly and a 4-step Nema17 gear reducer assembly.