UPDATE (05/Jun/2020): A subset of the dataset used in the BOP Challenge 2020, together with a new set of photorealistic training images generated by BlenderProc4BOP, can be downloaded from here.
UPDATE (09/Mar/2017): The format of the training and test data has been unified: 1) Folders obj_XX and scene_XX are now named XX, and 2) image-related information is saved in info.yml and the ground truth poses in gt.yml for both training and test images. The dataset in the previous format can be still downloaded from here.
We provide a python script that can be used to download any subset of the dataset. For instructions run:
python t-less_download.py --help
3D object models (10-40 MB per model set):
| CAD models | CAD models (subdivided) | Reconstructed models |
Training images - Primesense CARMINE 1.09 (~80 MB per object), [object id]:
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 |
Training images - Microsoft Kinect v2 (~90 MB per object), [object id]:
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 |
Training images - Canon IXUS 950 IS (~350 MB per object), [object id]:
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 |
Test images - Primesense CARMINE 1.09 (~410 MB per scene), [scene id]:
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 |
Test images - Microsoft Kinect v2 (~340 MB per scene), [scene id]:
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 |
Test images - Canon IXUS 950 IS (~1.4 GB per scene), [scene id]:
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 |