Unsupervised learning of object frames by dense equivariant image labelling
One of the key challenges of visual perception is to extract abstract models of 3D objects and object categories from visual measurements, which are affected by complex nuisance factors such as viewpoint, occlusion, motion, and deformations. Starting from the recent idea of viewpoint factorization,...
| Main Authors: | , , |
|---|---|
| Format: | Conference item |
| Language: | English |
| Published: |
Massachusetts Institute of Technology Press
2017
|
| _version_ | 1826264020968013824 |
|---|---|
| author | Thewlis, J Bilen, H Vedaldi, A |
| author_facet | Thewlis, J Bilen, H Vedaldi, A |
| author_sort | Thewlis, J |
| collection | OXFORD |
| description | One of the key challenges of visual perception is to extract abstract models of 3D objects and object categories from visual measurements, which are affected by complex nuisance factors such as viewpoint, occlusion, motion, and deformations. Starting from the recent idea of viewpoint factorization, we propose a new approach that, given a large number of images of an object and no other supervision, can extract a dense object-centric coordinate frame. This coordinate frame is invariant to deformations of the images and comes with a dense equivariant labelling neural network that can map image pixels to their corresponding object coordinates. We demonstrate the applicability of this method to simple articulated objects and deformable objects such as human faces, learning embeddings from random synthetic transformations or optical flow correspondences, all without any manual supervision. |
| first_indexed | 2024-03-06T20:01:08Z |
| format | Conference item |
| id | oxford-uuid:27544f49-8c3c-4d15-85a6-3ffdaba9d961 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-06T20:01:08Z |
| publishDate | 2017 |
| publisher | Massachusetts Institute of Technology Press |
| record_format | dspace |
| spelling | oxford-uuid:27544f49-8c3c-4d15-85a6-3ffdaba9d9612022-03-26T12:06:22ZUnsupervised learning of object frames by dense equivariant image labellingConference itemhttp://purl.org/coar/resource_type/c_5794uuid:27544f49-8c3c-4d15-85a6-3ffdaba9d961EnglishSymplectic Elements at OxfordMassachusetts Institute of Technology Press2017Thewlis, JBilen, HVedaldi, AOne of the key challenges of visual perception is to extract abstract models of 3D objects and object categories from visual measurements, which are affected by complex nuisance factors such as viewpoint, occlusion, motion, and deformations. Starting from the recent idea of viewpoint factorization, we propose a new approach that, given a large number of images of an object and no other supervision, can extract a dense object-centric coordinate frame. This coordinate frame is invariant to deformations of the images and comes with a dense equivariant labelling neural network that can map image pixels to their corresponding object coordinates. We demonstrate the applicability of this method to simple articulated objects and deformable objects such as human faces, learning embeddings from random synthetic transformations or optical flow correspondences, all without any manual supervision. |
| spellingShingle | Thewlis, J Bilen, H Vedaldi, A Unsupervised learning of object frames by dense equivariant image labelling |
| title | Unsupervised learning of object frames by dense equivariant image labelling |
| title_full | Unsupervised learning of object frames by dense equivariant image labelling |
| title_fullStr | Unsupervised learning of object frames by dense equivariant image labelling |
| title_full_unstemmed | Unsupervised learning of object frames by dense equivariant image labelling |
| title_short | Unsupervised learning of object frames by dense equivariant image labelling |
| title_sort | unsupervised learning of object frames by dense equivariant image labelling |
| work_keys_str_mv | AT thewlisj unsupervisedlearningofobjectframesbydenseequivariantimagelabelling AT bilenh unsupervisedlearningofobjectframesbydenseequivariantimagelabelling AT vedaldia unsupervisedlearningofobjectframesbydenseequivariantimagelabelling |