FlowNet3D++: Geometric losses for deep scene flow estimation
We present FlowNet3D++, a deep scene flow estimation network. Inspired by classical methods, FlowNet3D++ incorporates geometric constraints in the form of point-toplane distance and angular alignment between individual vectors in the flow field, into FlowNet3D [21]. We demonstrate that the addition...
| Prif Awduron: | , , , , |
|---|---|
| Fformat: | Conference item |
| Iaith: | English |
| Cyhoeddwyd: |
IEEE
2020
|
| _version_ | 1826293607432192000 |
|---|---|
| author | Wang, Z Li, S Howard-Jenkins, H Prisacariu, VA Chen, M |
| author_facet | Wang, Z Li, S Howard-Jenkins, H Prisacariu, VA Chen, M |
| author_sort | Wang, Z |
| collection | OXFORD |
| description | We present FlowNet3D++, a deep scene flow estimation network. Inspired by classical methods, FlowNet3D++ incorporates geometric constraints in the form of point-toplane distance and angular alignment between individual vectors in the flow field, into FlowNet3D [21]. We demonstrate that the addition of these geometric loss terms improves the previous state-of-art FlowNet3D accuracy from 57.85% to 63.43%. To further demonstrate the effectiveness of our geometric constraints, we propose a benchmark for flow estimation on the task of dynamic 3D reconstruction, thus providing a more holistic and practical measure of performance than the breakdown of individual metrics previously used to evaluate scene flow. This is made possible through the contribution of a novel pipeline to integrate point-based scene flow predictions into a global dense volume. FlowNet3D++ achieves up to a 15.0% reduction in reconstruction error over FlowNet3D, and up to a 35.2% improvement over KillingFusion [32] alone. We will release our scene flow estimation code later. |
| first_indexed | 2024-03-07T03:32:44Z |
| format | Conference item |
| id | oxford-uuid:bb4600c1-b7c8-431e-8d6c-1e72030f1b11 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-07T03:32:44Z |
| publishDate | 2020 |
| publisher | IEEE |
| record_format | dspace |
| spelling | oxford-uuid:bb4600c1-b7c8-431e-8d6c-1e72030f1b112022-03-27T05:15:45ZFlowNet3D++: Geometric losses for deep scene flow estimationConference itemhttp://purl.org/coar/resource_type/c_5794uuid:bb4600c1-b7c8-431e-8d6c-1e72030f1b11EnglishSymplectic ElementsIEEE2020Wang, ZLi, SHoward-Jenkins, HPrisacariu, VAChen, MWe present FlowNet3D++, a deep scene flow estimation network. Inspired by classical methods, FlowNet3D++ incorporates geometric constraints in the form of point-toplane distance and angular alignment between individual vectors in the flow field, into FlowNet3D [21]. We demonstrate that the addition of these geometric loss terms improves the previous state-of-art FlowNet3D accuracy from 57.85% to 63.43%. To further demonstrate the effectiveness of our geometric constraints, we propose a benchmark for flow estimation on the task of dynamic 3D reconstruction, thus providing a more holistic and practical measure of performance than the breakdown of individual metrics previously used to evaluate scene flow. This is made possible through the contribution of a novel pipeline to integrate point-based scene flow predictions into a global dense volume. FlowNet3D++ achieves up to a 15.0% reduction in reconstruction error over FlowNet3D, and up to a 35.2% improvement over KillingFusion [32] alone. We will release our scene flow estimation code later. |
| spellingShingle | Wang, Z Li, S Howard-Jenkins, H Prisacariu, VA Chen, M FlowNet3D++: Geometric losses for deep scene flow estimation |
| title | FlowNet3D++: Geometric losses for deep scene flow estimation |
| title_full | FlowNet3D++: Geometric losses for deep scene flow estimation |
| title_fullStr | FlowNet3D++: Geometric losses for deep scene flow estimation |
| title_full_unstemmed | FlowNet3D++: Geometric losses for deep scene flow estimation |
| title_short | FlowNet3D++: Geometric losses for deep scene flow estimation |
| title_sort | flownet3d geometric losses for deep scene flow estimation |
| work_keys_str_mv | AT wangz flownet3dgeometriclossesfordeepsceneflowestimation AT lis flownet3dgeometriclossesfordeepsceneflowestimation AT howardjenkinsh flownet3dgeometriclossesfordeepsceneflowestimation AT prisacariuva flownet3dgeometriclossesfordeepsceneflowestimation AT chenm flownet3dgeometriclossesfordeepsceneflowestimation |