SEM-GAT: explainable semantic pose estimation using learned graph attention
This paper proposes a Graph Neural Network (GNN)-based method for exploiting semantics and local geometry to guide the identification of reliable pointcloud registration candidates. Semantic and morphological features of the environment serve as key reference points for registration, enabling accura...
| Main Authors: | , , , , |
|---|---|
| Format: | Conference item |
| Language: | English |
| Published: |
IEEE
2024
|
| _version_ | 1797112426732191744 |
|---|---|
| author | Panagiotaki, E De Martini, D Pramatarov, G Gadd, M Kunze, L |
| author_facet | Panagiotaki, E De Martini, D Pramatarov, G Gadd, M Kunze, L |
| author_sort | Panagiotaki, E |
| collection | OXFORD |
| description | This paper proposes a Graph Neural Network
(GNN)-based method for exploiting semantics and local geometry to guide the identification of reliable pointcloud registration
candidates. Semantic and morphological features of the environment serve as key reference points for registration, enabling accurate lidar-based pose estimation. Our novel lightweight static
graph structure informs our attention-based node aggregation
network by identifying semantic-instance relationships, acting
as an inductive bias to significantly reduce the computational
burden of pointcloud registration. By connecting candidate
nodes and exploiting cross-graph attention, we identify confidence scores for all potential registration correspondences
and estimate the displacement between pointcloud scans. Our
pipeline enables introspective analysis of the model’s performance by correlating it with the individual contributions of
local structures in the environment, providing valuable insights
into the system’s behaviour. We test our method on the KITTI
odometry dataset, achieving competitive accuracy compared
to benchmark methods and a higher track smoothness while
relying on significantly fewer network parameters. |
| first_indexed | 2024-03-07T08:25:36Z |
| format | Conference item |
| id | oxford-uuid:09df9470-5b91-48e7-b3c3-f6cde131354c |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-07T08:25:36Z |
| publishDate | 2024 |
| publisher | IEEE |
| record_format | dspace |
| spelling | oxford-uuid:09df9470-5b91-48e7-b3c3-f6cde131354c2024-02-15T10:48:44ZSEM-GAT: explainable semantic pose estimation using learned graph attentionConference itemhttp://purl.org/coar/resource_type/c_5794uuid:09df9470-5b91-48e7-b3c3-f6cde131354cEnglishSymplectic ElementsIEEE2024Panagiotaki, EDe Martini, DPramatarov, GGadd, MKunze, LThis paper proposes a Graph Neural Network (GNN)-based method for exploiting semantics and local geometry to guide the identification of reliable pointcloud registration candidates. Semantic and morphological features of the environment serve as key reference points for registration, enabling accurate lidar-based pose estimation. Our novel lightweight static graph structure informs our attention-based node aggregation network by identifying semantic-instance relationships, acting as an inductive bias to significantly reduce the computational burden of pointcloud registration. By connecting candidate nodes and exploiting cross-graph attention, we identify confidence scores for all potential registration correspondences and estimate the displacement between pointcloud scans. Our pipeline enables introspective analysis of the model’s performance by correlating it with the individual contributions of local structures in the environment, providing valuable insights into the system’s behaviour. We test our method on the KITTI odometry dataset, achieving competitive accuracy compared to benchmark methods and a higher track smoothness while relying on significantly fewer network parameters. |
| spellingShingle | Panagiotaki, E De Martini, D Pramatarov, G Gadd, M Kunze, L SEM-GAT: explainable semantic pose estimation using learned graph attention |
| title | SEM-GAT: explainable semantic pose estimation using learned graph attention |
| title_full | SEM-GAT: explainable semantic pose estimation using learned graph attention |
| title_fullStr | SEM-GAT: explainable semantic pose estimation using learned graph attention |
| title_full_unstemmed | SEM-GAT: explainable semantic pose estimation using learned graph attention |
| title_short | SEM-GAT: explainable semantic pose estimation using learned graph attention |
| title_sort | sem gat explainable semantic pose estimation using learned graph attention |
| work_keys_str_mv | AT panagiotakie semgatexplainablesemanticposeestimationusinglearnedgraphattention AT demartinid semgatexplainablesemanticposeestimationusinglearnedgraphattention AT pramatarovg semgatexplainablesemanticposeestimationusinglearnedgraphattention AT gaddm semgatexplainablesemanticposeestimationusinglearnedgraphattention AT kunzel semgatexplainablesemanticposeestimationusinglearnedgraphattention |