NeuroMorph: unsupervised shape interpolation and correspondence in one go
We present NeuroMorph, a new neural network architecture that takes as input two 3D shapes and produces in one go, i.e. in a single feed forward pass, a smooth interpolation and point-to-point correspondences between them. The interpolation, expressed as a deformation field, changes the pose of the...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Conference item |
| Language: | English |
| Published: |
IEEE
2021
|
| _version_ | 1797064073006809088 |
|---|---|
| author | Eisenberger, M Novotny, D Kerchenbaum, G Labatut, P Neverova, N Cremers, D Vedaldi, A |
| author_facet | Eisenberger, M Novotny, D Kerchenbaum, G Labatut, P Neverova, N Cremers, D Vedaldi, A |
| author_sort | Eisenberger, M |
| collection | OXFORD |
| description | We present NeuroMorph, a new neural network architecture that takes as input two 3D shapes and produces in one go, i.e. in a single feed forward pass, a smooth interpolation and point-to-point correspondences between them. The interpolation, expressed as a deformation field, changes the pose of the source shape to resemble the target, but leaves the object identity unchanged. NeuroMorph uses an elegant architecture combining graph convolutions with global feature pooling to extract local features. During training, the model is incentivized to create realistic deformations by approximating geodesics on the underlying shape space manifold. This strong geometric prior allows to train our model end-to-end and in a fully unsupervised manner without requiring any manual correspondence annotations. NeuroMorph works well for a large variety of input shapes, including non-isometric pairs from different object categories. It obtains state-of-the-art results for both shape correspondence and interpolation tasks, matching or surpassing the performance of recent unsupervised and supervised methods on multiple benchmarks. |
| first_indexed | 2024-03-06T21:09:04Z |
| format | Conference item |
| id | oxford-uuid:3d83ccc4-3d38-4d04-94bc-2526fda172c7 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-06T21:09:04Z |
| publishDate | 2021 |
| publisher | IEEE |
| record_format | dspace |
| spelling | oxford-uuid:3d83ccc4-3d38-4d04-94bc-2526fda172c72022-03-26T14:19:51ZNeuroMorph: unsupervised shape interpolation and correspondence in one goConference itemhttp://purl.org/coar/resource_type/c_5794uuid:3d83ccc4-3d38-4d04-94bc-2526fda172c7EnglishSymplectic ElementsIEEE2021Eisenberger, MNovotny, DKerchenbaum, GLabatut, PNeverova, NCremers, DVedaldi, AWe present NeuroMorph, a new neural network architecture that takes as input two 3D shapes and produces in one go, i.e. in a single feed forward pass, a smooth interpolation and point-to-point correspondences between them. The interpolation, expressed as a deformation field, changes the pose of the source shape to resemble the target, but leaves the object identity unchanged. NeuroMorph uses an elegant architecture combining graph convolutions with global feature pooling to extract local features. During training, the model is incentivized to create realistic deformations by approximating geodesics on the underlying shape space manifold. This strong geometric prior allows to train our model end-to-end and in a fully unsupervised manner without requiring any manual correspondence annotations. NeuroMorph works well for a large variety of input shapes, including non-isometric pairs from different object categories. It obtains state-of-the-art results for both shape correspondence and interpolation tasks, matching or surpassing the performance of recent unsupervised and supervised methods on multiple benchmarks. |
| spellingShingle | Eisenberger, M Novotny, D Kerchenbaum, G Labatut, P Neverova, N Cremers, D Vedaldi, A NeuroMorph: unsupervised shape interpolation and correspondence in one go |
| title | NeuroMorph: unsupervised shape interpolation and correspondence in one go |
| title_full | NeuroMorph: unsupervised shape interpolation and correspondence in one go |
| title_fullStr | NeuroMorph: unsupervised shape interpolation and correspondence in one go |
| title_full_unstemmed | NeuroMorph: unsupervised shape interpolation and correspondence in one go |
| title_short | NeuroMorph: unsupervised shape interpolation and correspondence in one go |
| title_sort | neuromorph unsupervised shape interpolation and correspondence in one go |
| work_keys_str_mv | AT eisenbergerm neuromorphunsupervisedshapeinterpolationandcorrespondenceinonego AT novotnyd neuromorphunsupervisedshapeinterpolationandcorrespondenceinonego AT kerchenbaumg neuromorphunsupervisedshapeinterpolationandcorrespondenceinonego AT labatutp neuromorphunsupervisedshapeinterpolationandcorrespondenceinonego AT neverovan neuromorphunsupervisedshapeinterpolationandcorrespondenceinonego AT cremersd neuromorphunsupervisedshapeinterpolationandcorrespondenceinonego AT vedaldia neuromorphunsupervisedshapeinterpolationandcorrespondenceinonego |