Self-supervised learning of geometrically stable features through probabilistic introspection
Self-supervision can dramatically cut back the amount of manually-labeled data required to train deep neural networks. While self-supervision has usually been considered for tasks such as image classification, in this paper we aim at extending it to geometry-oriented tasks such as semantic matching...
| Egile Nagusiak: | , , , |
|---|---|
| Formatua: | Conference item |
| Argitaratua: |
Institute for Electrical and Electronics Engineers
2018
|
| _version_ | 1826292269815169024 |
|---|---|
| author | Novotny, D Albanie, S Larlus, D Vedaldi, A |
| author_facet | Novotny, D Albanie, S Larlus, D Vedaldi, A |
| author_sort | Novotny, D |
| collection | OXFORD |
| description | Self-supervision can dramatically cut back the amount of manually-labeled data required to train deep neural networks. While self-supervision has usually been considered for tasks such as image classification, in this paper we aim at extending it to geometry-oriented tasks such as semantic matching and part detection. We do so by building on several recent ideas in unsupervised landmark detection. Our approach learns dense distinctive visual descriptors from an unlabeled dataset of images using synthetic image transformations. It does so by means of a robust probabilistic formulation that can introspectively determine which image regions are likely to result in stable image matching. We show empirically that a network pre-trained in this manner requires significantly less supervision to learn semantic object parts compared to numerous pre-training alternatives. We also show that the pre-trained representation is excellent for semantic object matching. |
| first_indexed | 2024-03-07T03:12:05Z |
| format | Conference item |
| id | oxford-uuid:b488e633-ede6-4d39-a524-6f07e76914fc |
| institution | University of Oxford |
| last_indexed | 2024-03-07T03:12:05Z |
| publishDate | 2018 |
| publisher | Institute for Electrical and Electronics Engineers |
| record_format | dspace |
| spelling | oxford-uuid:b488e633-ede6-4d39-a524-6f07e76914fc2022-03-27T04:26:52ZSelf-supervised learning of geometrically stable features through probabilistic introspectionConference itemhttp://purl.org/coar/resource_type/c_5794uuid:b488e633-ede6-4d39-a524-6f07e76914fcSymplectic Elements at OxfordInstitute for Electrical and Electronics Engineers2018Novotny, DAlbanie, SLarlus, DVedaldi, ASelf-supervision can dramatically cut back the amount of manually-labeled data required to train deep neural networks. While self-supervision has usually been considered for tasks such as image classification, in this paper we aim at extending it to geometry-oriented tasks such as semantic matching and part detection. We do so by building on several recent ideas in unsupervised landmark detection. Our approach learns dense distinctive visual descriptors from an unlabeled dataset of images using synthetic image transformations. It does so by means of a robust probabilistic formulation that can introspectively determine which image regions are likely to result in stable image matching. We show empirically that a network pre-trained in this manner requires significantly less supervision to learn semantic object parts compared to numerous pre-training alternatives. We also show that the pre-trained representation is excellent for semantic object matching. |
| spellingShingle | Novotny, D Albanie, S Larlus, D Vedaldi, A Self-supervised learning of geometrically stable features through probabilistic introspection |
| title | Self-supervised learning of geometrically stable features through probabilistic introspection |
| title_full | Self-supervised learning of geometrically stable features through probabilistic introspection |
| title_fullStr | Self-supervised learning of geometrically stable features through probabilistic introspection |
| title_full_unstemmed | Self-supervised learning of geometrically stable features through probabilistic introspection |
| title_short | Self-supervised learning of geometrically stable features through probabilistic introspection |
| title_sort | self supervised learning of geometrically stable features through probabilistic introspection |
| work_keys_str_mv | AT novotnyd selfsupervisedlearningofgeometricallystablefeaturesthroughprobabilisticintrospection AT albanies selfsupervisedlearningofgeometricallystablefeaturesthroughprobabilisticintrospection AT larlusd selfsupervisedlearningofgeometricallystablefeaturesthroughprobabilisticintrospection AT vedaldia selfsupervisedlearningofgeometricallystablefeaturesthroughprobabilisticintrospection |