File Download
There are no files associated with this item.
Links for fulltext
(May Require Subscription)
- Publisher Website: 10.1109/TPAMI.2022.3227513
- Scopus: eid_2-s2.0-85144751278
- PMID: 37015515
- WOS: WOS:001004665900053
- Find via
Supplementary
- Citations:
- Appears in Collections:
Article: Open World Entity Segmentation
Title | Open World Entity Segmentation |
---|---|
Authors | |
Keywords | class-agnostic cross-dataset Image segmentation open-world |
Issue Date | 1-Jul-2023 |
Publisher | Institute of Electrical and Electronics Engineers |
Citation | IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023, v. 45, n. 7, p. 8743-8756 How to Cite? |
Abstract | We introduce a new image segmentation task, called Entity Segmentation (ES), which aims to segment all visual entities ( objects and stuffs) in an image without predicting their semantic labels. By removing the need of class label prediction, the models trained for such task can focus more on improving segmentation quality. It has many practical applications such as image manipulation and editing where the quality of segmentation masks is crucial but class labels are less important. We conduct the first-ever study to investigate the feasibility of convolutional center-based representation to segment things and stuffs in a unified manner, and show that such representation fits exceptionally well in the context of ES. More specifically, we propose a CondInst-like fully-convolutional architecture with two novel modules specifically designed to exploit the classagnostic and non-overlapping requirements of ES. Experiments show that the models designed and trained for ES significantly outperforms popular class-specific panoptic segmentation models in terms of segmentation quality. Moreover, an ES model can be easily trained on a combination of multiple datasets without the need to resolve label conflicts in dataset merging, and the model trained for ES on one or more datasets can generalize very well to other test datasets of unseen domains. The code has been released at https://github.com/dvlab-research/Entity. |
Persistent Identifier | http://hdl.handle.net/10722/331712 |
ISSN | 2023 Impact Factor: 20.8 2023 SCImago Journal Rankings: 6.158 |
ISI Accession Number ID |
DC Field | Value | Language |
---|---|---|
dc.contributor.author | Qi, L | - |
dc.contributor.author | Kuen, J | - |
dc.contributor.author | Wang, Y | - |
dc.contributor.author | Gu, JX | - |
dc.contributor.author | Zhao, HS | - |
dc.contributor.author | Torr, P | - |
dc.contributor.author | Lin, Z | - |
dc.contributor.author | Jia, JY | - |
dc.date.accessioned | 2023-09-21T06:58:14Z | - |
dc.date.available | 2023-09-21T06:58:14Z | - |
dc.date.issued | 2023-07-01 | - |
dc.identifier.citation | IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023, v. 45, n. 7, p. 8743-8756 | - |
dc.identifier.issn | 0162-8828 | - |
dc.identifier.uri | http://hdl.handle.net/10722/331712 | - |
dc.description.abstract | We introduce a new image segmentation task, called Entity Segmentation (ES), which aims to segment all visual entities ( objects and stuffs) in an image without predicting their semantic labels. By removing the need of class label prediction, the models trained for such task can focus more on improving segmentation quality. It has many practical applications such as image manipulation and editing where the quality of segmentation masks is crucial but class labels are less important. We conduct the first-ever study to investigate the feasibility of convolutional center-based representation to segment things and stuffs in a unified manner, and show that such representation fits exceptionally well in the context of ES. More specifically, we propose a CondInst-like fully-convolutional architecture with two novel modules specifically designed to exploit the classagnostic and non-overlapping requirements of ES. Experiments show that the models designed and trained for ES significantly outperforms popular class-specific panoptic segmentation models in terms of segmentation quality. Moreover, an ES model can be easily trained on a combination of multiple datasets without the need to resolve label conflicts in dataset merging, and the model trained for ES on one or more datasets can generalize very well to other test datasets of unseen domains. The code has been released at https://github.com/dvlab-research/Entity. | - |
dc.language | eng | - |
dc.publisher | Institute of Electrical and Electronics Engineers | - |
dc.relation.ispartof | IEEE Transactions on Pattern Analysis and Machine Intelligence | - |
dc.rights | This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License. | - |
dc.subject | class-agnostic | - |
dc.subject | cross-dataset | - |
dc.subject | Image segmentation | - |
dc.subject | open-world | - |
dc.title | Open World Entity Segmentation | - |
dc.type | Article | - |
dc.identifier.doi | 10.1109/TPAMI.2022.3227513 | - |
dc.identifier.pmid | 37015515 | - |
dc.identifier.scopus | eid_2-s2.0-85144751278 | - |
dc.identifier.volume | 45 | - |
dc.identifier.issue | 7 | - |
dc.identifier.spage | 8743 | - |
dc.identifier.epage | 8756 | - |
dc.identifier.eissn | 1939-3539 | - |
dc.identifier.isi | WOS:001004665900053 | - |
dc.publisher.place | LOS ALAMITOS | - |
dc.identifier.issnl | 0162-8828 | - |