GDN: A Coarse-To-Fine (C2F) Representation for End-To-End 6-DoF Grasp Detection

Kuang-Yu Jeng, Yueh-Cheng Liu, Zhe Yu Liu, Jen-Wei Wang, Ya-Liang Chang, Hung-Ting Su, Winston Hsu
Proceedings of the 2020 Conference on Robot Learning, PMLR 155:220-231, 2021.

Abstract

We proposed an end-to-end grasp detection network, Grasp Detection Network (GDN), cooperated with a novel coarse-to-fine (C2F) grasp representation design to detect diverse and accurate 6-DoF grasps based on point clouds. Compared to previous two-stage approaches which sample and evaluate multiple grasp candidates, our architecture is at least 20 times faster. It is also 8% and 40% more accurate in terms of the success rate in single object scenes and the complete rate in clutter scenes, respectively. Our method shows superior results among settings with different number of views and input points. Moreover, we propose a new AP-based metric which considers both rotation and transition errors, making it a more comprehensive evaluation tool for grasp detection models.

Cite this Paper

BibTeX
@InProceedings{pmlr-v155-jeng21a, title = {GDN: A Coarse-To-Fine (C2F) Representation for End-To-End 6-DoF Grasp Detection}, author = {Jeng, Kuang-Yu and Liu, Yueh-Cheng and Liu, Zhe Yu and Wang, Jen-Wei and Chang, Ya-Liang and Su, Hung-Ting and Hsu, Winston}, booktitle = {Proceedings of the 2020 Conference on Robot Learning}, pages = {220--231}, year = {2021}, editor = {Kober, Jens and Ramos, Fabio and Tomlin, Claire}, volume = {155}, series = {Proceedings of Machine Learning Research}, month = {16--18 Nov}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v155/jeng21a/jeng21a.pdf}, url = {https://proceedings.mlr.press/v155/jeng21a.html}, abstract = {We proposed an end-to-end grasp detection network, Grasp Detection Network (GDN), cooperated with a novel coarse-to-fine (C2F) grasp representation design to detect diverse and accurate 6-DoF grasps based on point clouds. Compared to previous two-stage approaches which sample and evaluate multiple grasp candidates, our architecture is at least 20 times faster. It is also 8% and 40% more accurate in terms of the success rate in single object scenes and the complete rate in clutter scenes, respectively. Our method shows superior results among settings with different number of views and input points. Moreover, we propose a new AP-based metric which considers both rotation and transition errors, making it a more comprehensive evaluation tool for grasp detection models.} }
Endnote
%0 Conference Paper %T GDN: A Coarse-To-Fine (C2F) Representation for End-To-End 6-DoF Grasp Detection %A Kuang-Yu Jeng %A Yueh-Cheng Liu %A Zhe Yu Liu %A Jen-Wei Wang %A Ya-Liang Chang %A Hung-Ting Su %A Winston Hsu %B Proceedings of the 2020 Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2021 %E Jens Kober %E Fabio Ramos %E Claire Tomlin %F pmlr-v155-jeng21a %I PMLR %P 220--231 %U https://proceedings.mlr.press/v155/jeng21a.html %V 155 %X We proposed an end-to-end grasp detection network, Grasp Detection Network (GDN), cooperated with a novel coarse-to-fine (C2F) grasp representation design to detect diverse and accurate 6-DoF grasps based on point clouds. Compared to previous two-stage approaches which sample and evaluate multiple grasp candidates, our architecture is at least 20 times faster. It is also 8% and 40% more accurate in terms of the success rate in single object scenes and the complete rate in clutter scenes, respectively. Our method shows superior results among settings with different number of views and input points. Moreover, we propose a new AP-based metric which considers both rotation and transition errors, making it a more comprehensive evaluation tool for grasp detection models.
APA
Jeng, K., Liu, Y., Liu, Z.Y., Wang, J., Chang, Y., Su, H. & Hsu, W.. (2021). GDN: A Coarse-To-Fine (C2F) Representation for End-To-End 6-DoF Grasp Detection. Proceedings of the 2020 Conference on Robot Learning, in Proceedings of Machine Learning Research 155:220-231 Available from https://proceedings.mlr.press/v155/jeng21a.html.

Related Material