Visuo-tactile recognition of partial point clouds using PointNet and curriculum learning: enabling tactile perception from visual data
This article is about recognizing handheld objects from incomplete tactile observations with a classifier trained on only visual representations. Our method is based on the deep learning (DL) architecture PointNet and a curriculum learning (CL) technique for fostering the learning of descriptors rob...
| Main Authors: | , , , |
|---|---|
| 格式: | Journal article |
| 語言: | English |
| 出版: |
IEEE
2022
|
| _version_ | 1826309479255244800 |
|---|---|
| author | Parsons, C Albini, A Martini, DD Maiolino, P |
| author_facet | Parsons, C Albini, A Martini, DD Maiolino, P |
| author_sort | Parsons, C |
| collection | OXFORD |
| description | This article is about recognizing handheld objects from incomplete tactile observations with a classifier trained on only visual representations. Our method is based on the deep learning (DL) architecture PointNet and a curriculum learning (CL) technique for fostering the learning of descriptors robust to partial representations of objects. The learning procedure gradually decomposes the visual point clouds to synthesize sparser and sparser input data for the model. In this manner, we were able to employ one-shot learning, using the decomposed visual point clouds as augmentations, and reduce the data-collection requirement for training. The approach allows for a gradual improvement of prediction accuracy as more tactile data become available. |
| first_indexed | 2024-03-07T07:34:53Z |
| format | Journal article |
| id | oxford-uuid:68100d51-ee5a-43bc-9890-514e49098da3 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-07T07:34:53Z |
| publishDate | 2022 |
| publisher | IEEE |
| record_format | dspace |
| spelling | oxford-uuid:68100d51-ee5a-43bc-9890-514e49098da32023-03-06T15:49:53ZVisuo-tactile recognition of partial point clouds using PointNet and curriculum learning: enabling tactile perception from visual dataJournal articlehttp://purl.org/coar/resource_type/c_dcae04bcuuid:68100d51-ee5a-43bc-9890-514e49098da3EnglishSymplectic ElementsIEEE2022Parsons, CAlbini, AMartini, DDMaiolino, PThis article is about recognizing handheld objects from incomplete tactile observations with a classifier trained on only visual representations. Our method is based on the deep learning (DL) architecture PointNet and a curriculum learning (CL) technique for fostering the learning of descriptors robust to partial representations of objects. The learning procedure gradually decomposes the visual point clouds to synthesize sparser and sparser input data for the model. In this manner, we were able to employ one-shot learning, using the decomposed visual point clouds as augmentations, and reduce the data-collection requirement for training. The approach allows for a gradual improvement of prediction accuracy as more tactile data become available. |
| spellingShingle | Parsons, C Albini, A Martini, DD Maiolino, P Visuo-tactile recognition of partial point clouds using PointNet and curriculum learning: enabling tactile perception from visual data |
| title | Visuo-tactile recognition of partial point clouds using PointNet and curriculum learning: enabling tactile perception from visual data |
| title_full | Visuo-tactile recognition of partial point clouds using PointNet and curriculum learning: enabling tactile perception from visual data |
| title_fullStr | Visuo-tactile recognition of partial point clouds using PointNet and curriculum learning: enabling tactile perception from visual data |
| title_full_unstemmed | Visuo-tactile recognition of partial point clouds using PointNet and curriculum learning: enabling tactile perception from visual data |
| title_short | Visuo-tactile recognition of partial point clouds using PointNet and curriculum learning: enabling tactile perception from visual data |
| title_sort | visuo tactile recognition of partial point clouds using pointnet and curriculum learning enabling tactile perception from visual data |
| work_keys_str_mv | AT parsonsc visuotactilerecognitionofpartialpointcloudsusingpointnetandcurriculumlearningenablingtactileperceptionfromvisualdata AT albinia visuotactilerecognitionofpartialpointcloudsusingpointnetandcurriculumlearningenablingtactileperceptionfromvisualdata AT martinidd visuotactilerecognitionofpartialpointcloudsusingpointnetandcurriculumlearningenablingtactileperceptionfromvisualdata AT maiolinop visuotactilerecognitionofpartialpointcloudsusingpointnetandcurriculumlearningenablingtactileperceptionfromvisualdata |