Contrastive Learning for 3D Point Clouds Classification and Shape Completion
In this paper, we present the idea of Self Supervised learning on the shape completion and classification of point clouds. Most 3D shape completion pipelines utilize AutoEncoders to extract features from point clouds used in downstream tasks such as classification, segmentation, detection, and other...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2021-11-01
|
| Series: | Sensors |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1424-8220/21/21/7392 |
| _version_ | 1797511682863398912 |
|---|---|
| author | Danish Nazir Muhammad Zeshan Afzal Alain Pagani Marcus Liwicki Didier Stricker |
| author_facet | Danish Nazir Muhammad Zeshan Afzal Alain Pagani Marcus Liwicki Didier Stricker |
| author_sort | Danish Nazir |
| collection | DOAJ |
| description | In this paper, we present the idea of Self Supervised learning on the shape completion and classification of point clouds. Most 3D shape completion pipelines utilize AutoEncoders to extract features from point clouds used in downstream tasks such as classification, segmentation, detection, and other related applications. Our idea is to add contrastive learning into AutoEncoders to encourage global feature learning of the point cloud classes. It is performed by optimizing triplet loss. Furthermore, local feature representations learning of point cloud is performed by adding the Chamfer distance function. To evaluate the performance of our approach, we utilize the PointNet classifier. We also extend the number of classes for evaluation from 4 to 10 to show the generalization ability of the learned features. Based on our results, embeddings generated from the contrastive AutoEncoder enhances shape completion and classification performance from <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>84.2</mn><mo>%</mo></mrow></semantics></math></inline-formula> to <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>84.9</mn><mo>%</mo></mrow></semantics></math></inline-formula> of point clouds achieving the state-of-the-art results with 10 classes. |
| first_indexed | 2024-03-10T05:51:34Z |
| format | Article |
| id | doaj.art-0692d70bba7b47738add93318f10650f |
| institution | Directory Open Access Journal |
| issn | 1424-8220 |
| language | English |
| last_indexed | 2024-03-10T05:51:34Z |
| publishDate | 2021-11-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Sensors |
| spelling | doaj.art-0692d70bba7b47738add93318f10650f2023-11-22T21:40:56ZengMDPI AGSensors1424-82202021-11-012121739210.3390/s21217392Contrastive Learning for 3D Point Clouds Classification and Shape CompletionDanish Nazir0Muhammad Zeshan Afzal1Alain Pagani2Marcus Liwicki3Didier Stricker4Department of Computer Science, Technical University of Kaiserslautern, 67663 Kaiserslautern, GermanyDepartment of Computer Science, Technical University of Kaiserslautern, 67663 Kaiserslautern, GermanyGerman Research Institute for Artificial Intelligence (DFKI), 67663 Kaiserslautern, GermanyDepartment of Computer Science, Luleå University of Technology, 971 87 Luleå, SwedenDepartment of Computer Science, Technical University of Kaiserslautern, 67663 Kaiserslautern, GermanyIn this paper, we present the idea of Self Supervised learning on the shape completion and classification of point clouds. Most 3D shape completion pipelines utilize AutoEncoders to extract features from point clouds used in downstream tasks such as classification, segmentation, detection, and other related applications. Our idea is to add contrastive learning into AutoEncoders to encourage global feature learning of the point cloud classes. It is performed by optimizing triplet loss. Furthermore, local feature representations learning of point cloud is performed by adding the Chamfer distance function. To evaluate the performance of our approach, we utilize the PointNet classifier. We also extend the number of classes for evaluation from 4 to 10 to show the generalization ability of the learned features. Based on our results, embeddings generated from the contrastive AutoEncoder enhances shape completion and classification performance from <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>84.2</mn><mo>%</mo></mrow></semantics></math></inline-formula> to <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>84.9</mn><mo>%</mo></mrow></semantics></math></inline-formula> of point clouds achieving the state-of-the-art results with 10 classes.https://www.mdpi.com/1424-8220/21/21/7392point cloud classificationpoint cloud shape completionAutoEncoderscontrastive AutoEncoderscontrasitive learning for point cloudsself-supervised learning for point cloud shape completion |
| spellingShingle | Danish Nazir Muhammad Zeshan Afzal Alain Pagani Marcus Liwicki Didier Stricker Contrastive Learning for 3D Point Clouds Classification and Shape Completion Sensors point cloud classification point cloud shape completion AutoEncoders contrastive AutoEncoders contrasitive learning for point clouds self-supervised learning for point cloud shape completion |
| title | Contrastive Learning for 3D Point Clouds Classification and Shape Completion |
| title_full | Contrastive Learning for 3D Point Clouds Classification and Shape Completion |
| title_fullStr | Contrastive Learning for 3D Point Clouds Classification and Shape Completion |
| title_full_unstemmed | Contrastive Learning for 3D Point Clouds Classification and Shape Completion |
| title_short | Contrastive Learning for 3D Point Clouds Classification and Shape Completion |
| title_sort | contrastive learning for 3d point clouds classification and shape completion |
| topic | point cloud classification point cloud shape completion AutoEncoders contrastive AutoEncoders contrasitive learning for point clouds self-supervised learning for point cloud shape completion |
| url | https://www.mdpi.com/1424-8220/21/21/7392 |
| work_keys_str_mv | AT danishnazir contrastivelearningfor3dpointcloudsclassificationandshapecompletion AT muhammadzeshanafzal contrastivelearningfor3dpointcloudsclassificationandshapecompletion AT alainpagani contrastivelearningfor3dpointcloudsclassificationandshapecompletion AT marcusliwicki contrastivelearningfor3dpointcloudsclassificationandshapecompletion AT didierstricker contrastivelearningfor3dpointcloudsclassificationandshapecompletion |