Learning 3D object categories by looking around them
Traditional approaches for learning 3D object categories use either synthetic data or manual supervision. In this paper, we propose a method which does not require manual annotations and is instead cued by observing objects from a moving vantage point. Our system builds on two innovations: a Siamese...
| Hlavní autoři: | , , |
|---|---|
| Médium: | Conference item |
| Jazyk: | English |
| Vydáno: |
IEEE
2017
|
| _version_ | 1826271773201530880 |
|---|---|
| author | Novotny, D Larlus, D Vedaldi, A |
| author_facet | Novotny, D Larlus, D Vedaldi, A |
| author_sort | Novotny, D |
| collection | OXFORD |
| description | Traditional approaches for learning 3D object categories use either synthetic data or manual supervision. In this paper, we propose a method which does not require manual annotations and is instead cued by observing objects from a moving vantage point. Our system builds on two innovations: a Siamese viewpoint factorization network that robustly aligns different videos together without explicitly comparing 3D shapes; and a 3D shape completion network that can extract the full shape of an object from partial observations. We also demonstrate the benefits of configuring networks to perform probabilistic predictions as well as of geometry-aware data augmentation schemes. We obtain state-of-the-art results on publicly-available benchmarks. |
| first_indexed | 2024-03-06T22:02:00Z |
| format | Conference item |
| id | oxford-uuid:4eebe356-a6fd-49f1-97d5-5cbcdd36f99b |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-06T22:02:00Z |
| publishDate | 2017 |
| publisher | IEEE |
| record_format | dspace |
| spelling | oxford-uuid:4eebe356-a6fd-49f1-97d5-5cbcdd36f99b2022-03-26T16:04:01ZLearning 3D object categories by looking around themConference itemhttp://purl.org/coar/resource_type/c_5794uuid:4eebe356-a6fd-49f1-97d5-5cbcdd36f99bEnglishSymplectic Elements at OxfordIEEE2017Novotny, DLarlus, DVedaldi, ATraditional approaches for learning 3D object categories use either synthetic data or manual supervision. In this paper, we propose a method which does not require manual annotations and is instead cued by observing objects from a moving vantage point. Our system builds on two innovations: a Siamese viewpoint factorization network that robustly aligns different videos together without explicitly comparing 3D shapes; and a 3D shape completion network that can extract the full shape of an object from partial observations. We also demonstrate the benefits of configuring networks to perform probabilistic predictions as well as of geometry-aware data augmentation schemes. We obtain state-of-the-art results on publicly-available benchmarks. |
| spellingShingle | Novotny, D Larlus, D Vedaldi, A Learning 3D object categories by looking around them |
| title | Learning 3D object categories by looking around them |
| title_full | Learning 3D object categories by looking around them |
| title_fullStr | Learning 3D object categories by looking around them |
| title_full_unstemmed | Learning 3D object categories by looking around them |
| title_short | Learning 3D object categories by looking around them |
| title_sort | learning 3d object categories by looking around them |
| work_keys_str_mv | AT novotnyd learning3dobjectcategoriesbylookingaroundthem AT larlusd learning3dobjectcategoriesbylookingaroundthem AT vedaldia learning3dobjectcategoriesbylookingaroundthem |