DC Field | Value | Language |
---|---|---|
dc.contributor.author | 박충현 | - |
dc.date.accessioned | 2023-04-07T16:32:56Z | - |
dc.date.available | 2023-04-07T16:32:56Z | - |
dc.date.issued | 2022 | - |
dc.identifier.other | OAK-2015-09765 | - |
dc.identifier.uri | http://postech.dcollection.net/common/orgView/200000600797 | ko_KR |
dc.identifier.uri | https://oasis.postech.ac.kr/handle/2014.oak/117219 | - |
dc.description | Master | - |
dc.description.abstract | The recent success of neural networks enables a better interpretation of 3D point clouds, but processing a large-scale 3D scene remains a challenging problem. Most current approaches divide a large-scale scene into small regions and combine the local predictions together. However, this scheme inevitably involves additional stages for pre- and post-processing and may also degrade the final output due to predictions in a local perspective. This thesis introduces Fast Point Transformer that consists of a novel lightweight self-attention layer. Our approach encodes continuous 3D coordinates, and the voxel hashing-based architecture boosts computational efficiency. The proposed method is demonstrated with 3D semantic segmentation and 3D detection. The accuracy of our approach is competitive to the best voxel-based method, and our network achieves 136 times faster inference time than the state-of-the-art, Point Transformer, with a reasonable accuracy trade-off. | - |
dc.language | eng | - |
dc.publisher | 포항공과대학교 | - |
dc.title | Fast Point Transformer for Large-scale 3D Scene Understanding | - |
dc.title.alternative | 대규모 3차원 장면 이해를 위한 빠른 포인트 트랜스포머 | - |
dc.type | Thesis | - |
dc.contributor.college | 인공지능대학원 | - |
dc.date.degree | 2022- 2 | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.
library@postech.ac.kr Tel: 054-279-2548
Copyrights © by 2017 Pohang University of Science ad Technology All right reserved.