D-Textureless dataset

We provide a dataset which includes 9 texture-less models (used for training) and 55 test scenes with clutter and occlusions. It has been acquired with a webcam and comes with hand-labeled groundtruth for the pose of each model instance in the scene. Each scene includes one or more models, but one instance of each at most.

This dataset has been proposed in [1]. If you plan to use this dataset for your research activity, please reference the publication accordingly.

[download]

train_set (MODELS)

test_set_sample02 test_set_sample01 (2 SAMPLE SCENES)

Ground truth file structure (model indices are 0-based):

N
model index #1
model index #2
...
model index #N

A Line2D training file is also included in the “train” directory. It can be loaded through OpenCV by calling:

cv::linemod::Detector::read()

Note that Line2D is available in OpenCV from version 2.4 onwards.

 

References:

[1] F. Tombari, A. Franchi, L. Di Stefano, “BOLD features to detect texture-less objects”, ICCV 2013