An intuitive multi-frequency feature representation for SO(3)-equivariant networks
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866910437977620480 |
|---|---|
| author | Son, Dongwon Kim, Jaehyung Son, Sanghyeon Kim, Beomjoon |
| author_facet | Son, Dongwon Kim, Jaehyung Son, Sanghyeon Kim, Beomjoon |
| contents | The usage of 3D vision algorithms, such as shape reconstruction, remains limited because they require inputs to be at a fixed canonical rotation. Recently, a simple equivariant network, Vector Neuron (VN) has been proposed that can be easily used with the state-of-the-art 3D neural network (NN) architectures. However, its performance is limited because it is designed to use only three-dimensional features, which is insufficient to capture the details present in 3D data. In this paper, we introduce an equivariant feature representation for mapping a 3D point to a high-dimensional feature space. Our feature can discern multiple frequencies present in 3D data, which is the key to designing an expressive feature for 3D vision tasks. Our representation can be used as an input to VNs, and the results demonstrate that with our feature representation, VN captures more details, overcoming the limitation raised in its original paper. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2405_04537 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | An intuitive multi-frequency feature representation for SO(3)-equivariant networks Son, Dongwon Kim, Jaehyung Son, Sanghyeon Kim, Beomjoon Computer Vision and Pattern Recognition Artificial Intelligence Graphics The usage of 3D vision algorithms, such as shape reconstruction, remains limited because they require inputs to be at a fixed canonical rotation. Recently, a simple equivariant network, Vector Neuron (VN) has been proposed that can be easily used with the state-of-the-art 3D neural network (NN) architectures. However, its performance is limited because it is designed to use only three-dimensional features, which is insufficient to capture the details present in 3D data. In this paper, we introduce an equivariant feature representation for mapping a 3D point to a high-dimensional feature space. Our feature can discern multiple frequencies present in 3D data, which is the key to designing an expressive feature for 3D vision tasks. Our representation can be used as an input to VNs, and the results demonstrate that with our feature representation, VN captures more details, overcoming the limitation raised in its original paper. |
| title | An intuitive multi-frequency feature representation for SO(3)-equivariant networks |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence Graphics |
| url | https://arxiv.org/abs/2405.04537 |