Recurrent instance segmentation
Instance segmentation is the problem of detecting and delineating each distinct object of interest appearing in an image. Current instance segmentation approaches consist of ensembles of modules that are trained independently of each other, thus missing opportunities for joint learning. Here we prop...
| Main Authors: | , |
|---|---|
| Format: | Conference item |
| Published: |
Springer Verlag
2016
|
| _version_ | 1826284599352754176 |
|---|---|
| author | Romera-Paredes, B Torr, P |
| author_facet | Romera-Paredes, B Torr, P |
| author_sort | Romera-Paredes, B |
| collection | OXFORD |
| description | Instance segmentation is the problem of detecting and delineating each distinct object of interest appearing in an image. Current instance segmentation approaches consist of ensembles of modules that are trained independently of each other, thus missing opportunities for joint learning. Here we propose a new instance segmentation paradigm consisting in an end-to-end method that learns how to segment instances sequentially. The model is based on a recurrent neural network that sequentially finds objects and their segmentations one at a time. This net is provided with a spatial memory that keeps track of what pixels have been explained and allows occlusion handling. In order to train the model we designed a principled loss function that accurately represents the properties of the instance segmentation problem. In the experiments carried out, we found that our method outperforms recent approaches on multiple person segmentation, and all state of the art approaches on the Plant Phenotyping dataset for leaf counting. |
| first_indexed | 2024-03-07T01:16:16Z |
| format | Conference item |
| id | oxford-uuid:8ecd16f3-51b2-4708-aa70-aa6aa2d75f31 |
| institution | University of Oxford |
| last_indexed | 2024-03-07T01:16:16Z |
| publishDate | 2016 |
| publisher | Springer Verlag |
| record_format | dspace |
| spelling | oxford-uuid:8ecd16f3-51b2-4708-aa70-aa6aa2d75f312022-03-26T23:00:01ZRecurrent instance segmentationConference itemhttp://purl.org/coar/resource_type/c_5794uuid:8ecd16f3-51b2-4708-aa70-aa6aa2d75f31Symplectic Elements at OxfordSpringer Verlag2016Romera-Paredes, BTorr, PInstance segmentation is the problem of detecting and delineating each distinct object of interest appearing in an image. Current instance segmentation approaches consist of ensembles of modules that are trained independently of each other, thus missing opportunities for joint learning. Here we propose a new instance segmentation paradigm consisting in an end-to-end method that learns how to segment instances sequentially. The model is based on a recurrent neural network that sequentially finds objects and their segmentations one at a time. This net is provided with a spatial memory that keeps track of what pixels have been explained and allows occlusion handling. In order to train the model we designed a principled loss function that accurately represents the properties of the instance segmentation problem. In the experiments carried out, we found that our method outperforms recent approaches on multiple person segmentation, and all state of the art approaches on the Plant Phenotyping dataset for leaf counting. |
| spellingShingle | Romera-Paredes, B Torr, P Recurrent instance segmentation |
| title | Recurrent instance segmentation |
| title_full | Recurrent instance segmentation |
| title_fullStr | Recurrent instance segmentation |
| title_full_unstemmed | Recurrent instance segmentation |
| title_short | Recurrent instance segmentation |
| title_sort | recurrent instance segmentation |
| work_keys_str_mv | AT romeraparedesb recurrentinstancesegmentation AT torrp recurrentinstancesegmentation |