Topology Prediction of Branched Deformable Linear Objects Using Deep Learning
Automated wire harness handling can improve production efficiency, increase quality, and reduce assembly costs. However, due to deformation, there are an infinite number of possible wire harness configurations, making wire harness perception a challenge. Deep learning is a popular method for compute...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2024-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10804119/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1846102142156275712 |
|---|---|
| author | Shengzhe Ouyang Manuel Zurn Lukas Zeh Armin Lechler Alexander Verl |
| author_facet | Shengzhe Ouyang Manuel Zurn Lukas Zeh Armin Lechler Alexander Verl |
| author_sort | Shengzhe Ouyang |
| collection | DOAJ |
| description | Automated wire harness handling can improve production efficiency, increase quality, and reduce assembly costs. However, due to deformation, there are an infinite number of possible wire harness configurations, making wire harness perception a challenge. Deep learning is a popular method for computer vision but lacks datasets, models, and experiments for wire harness perception. Therefore, this paper presents a novel deep learning model to predict the configuration of a wire harness using artificially generated datasets mixed with real annotated data. The model predicts keypoints which are interpolated as cubic splines to represent the wire harness configuration with reduced degrees of freedom. We benchmark our novel model against YOLOv8-Pose and experiment with different possibilities for predicting the wire harness. As a result, our proposed approach achieves mAP@50-95 of 89.8%, which could further be integrated into robotic systems to improve the automation and precision of robotic wire harness handling. |
| format | Article |
| id | doaj-art-826d881fbc4146d196960142303cf681 |
| institution | Kabale University |
| issn | 2169-3536 |
| language | English |
| publishDate | 2024-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj-art-826d881fbc4146d196960142303cf6812024-12-28T00:00:54ZengIEEEIEEE Access2169-35362024-01-011219439919441110.1109/ACCESS.2024.351863410804119Topology Prediction of Branched Deformable Linear Objects Using Deep LearningShengzhe Ouyang0https://orcid.org/0009-0007-7320-1996Manuel Zurn1https://orcid.org/0000-0002-0409-5540Lukas Zeh2https://orcid.org/0000-0003-2730-1383Armin Lechler3https://orcid.org/0000-0002-4073-1487Alexander Verl4https://orcid.org/0000-0002-2548-6620Institute for Control Engineering and Manufacturing Units, University of Stuttgart, Stuttgart, GermanyInstitute for Control Engineering and Manufacturing Units, University of Stuttgart, Stuttgart, GermanyInstitute for Control Engineering and Manufacturing Units, University of Stuttgart, Stuttgart, GermanyInstitute for Control Engineering and Manufacturing Units, University of Stuttgart, Stuttgart, GermanyInstitute for Control Engineering and Manufacturing Units, University of Stuttgart, Stuttgart, GermanyAutomated wire harness handling can improve production efficiency, increase quality, and reduce assembly costs. However, due to deformation, there are an infinite number of possible wire harness configurations, making wire harness perception a challenge. Deep learning is a popular method for computer vision but lacks datasets, models, and experiments for wire harness perception. Therefore, this paper presents a novel deep learning model to predict the configuration of a wire harness using artificially generated datasets mixed with real annotated data. The model predicts keypoints which are interpolated as cubic splines to represent the wire harness configuration with reduced degrees of freedom. We benchmark our novel model against YOLOv8-Pose and experiment with different possibilities for predicting the wire harness. As a result, our proposed approach achieves mAP@50-95 of 89.8%, which could further be integrated into robotic systems to improve the automation and precision of robotic wire harness handling.https://ieeexplore.ieee.org/document/10804119/Machine visionartificial intelligencedeep learningsynthetic datasettransfer learningbranched deformable linear objects |
| spellingShingle | Shengzhe Ouyang Manuel Zurn Lukas Zeh Armin Lechler Alexander Verl Topology Prediction of Branched Deformable Linear Objects Using Deep Learning IEEE Access Machine vision artificial intelligence deep learning synthetic dataset transfer learning branched deformable linear objects |
| title | Topology Prediction of Branched Deformable Linear Objects Using Deep Learning |
| title_full | Topology Prediction of Branched Deformable Linear Objects Using Deep Learning |
| title_fullStr | Topology Prediction of Branched Deformable Linear Objects Using Deep Learning |
| title_full_unstemmed | Topology Prediction of Branched Deformable Linear Objects Using Deep Learning |
| title_short | Topology Prediction of Branched Deformable Linear Objects Using Deep Learning |
| title_sort | topology prediction of branched deformable linear objects using deep learning |
| topic | Machine vision artificial intelligence deep learning synthetic dataset transfer learning branched deformable linear objects |
| url | https://ieeexplore.ieee.org/document/10804119/ |
| work_keys_str_mv | AT shengzheouyang topologypredictionofbrancheddeformablelinearobjectsusingdeeplearning AT manuelzurn topologypredictionofbrancheddeformablelinearobjectsusingdeeplearning AT lukaszeh topologypredictionofbrancheddeformablelinearobjectsusingdeeplearning AT arminlechler topologypredictionofbrancheddeformablelinearobjectsusingdeeplearning AT alexanderverl topologypredictionofbrancheddeformablelinearobjectsusingdeeplearning |