Open-world instance segmentation: Exploiting pseudo ground truth from learned pairwise affinity

W Wang, M Feiszli, H Wang… - Proceedings of the …, 2022 - openaccess.thecvf.com
Open-world instance segmentation is the task of grouping pixels into object instances
without any pre-determined taxonomy. This is challenging, as state-of-the-art methods rely …
被引用次数：50 相关文章
[PDF] thecvf.com

[PDF][PDF] Open-World Instance Segmentation: Exploiting Pseudo Ground Truth From Learned Pairwise Affinity

W Wang, M Feiszli, H Wang, J Malik, D Tran - openaccess.thecvf.com
In section 5 of main paper, we showed how our proposed method generates pseudo-GT
masks on unlabeled images, and how GGN benefited from training on unlabeled images …

Open-World Instance Segmentation: Exploiting Pseudo Ground Truth From Learned Pairwise Affinity

W Wang, M Feiszli, H Wang, J Malik… - 2022 IEEE/CVF …, 2022 - ieeexplore.ieee.org
Open-world instance segmentation is the task of grouping pixels into object instances
without any pre-determined taxonomy. This is challenging, as state-of-the-art methods rely …

Open-World Instance Segmentation: Exploiting Pseudo Ground Truth From Learned Pairwise Affinity

W Wang, M Feiszli, H Wang, J Malik, D Tran - arXiv preprint arXiv …, 2022 - arxiv.org
Open-world instance segmentation is the task of grouping pixels into object instances
without any pre-determined taxonomy. This is challenging, as state-of-the-art methods rely …

Open-World Instance Segmentation: Exploiting Pseudo Ground Truth From Learned Pairwise Affinity

W Wang, M Feiszli, H Wang, J Malik… - 2022 IEEE/CVF …, 2022 - computer.org
Open-world instance segmentation is the task of grouping pixels into object instances
without any pre-determined taxonomy. This is challenging, as state-of-the-art methods rely …

Open-World Instance Segmentation: Exploiting Pseudo Ground Truth From Learned Pairwise Affinity

W Wang, M Feiszli, H Wang, J Malik, D Tran - arXiv e-prints, 2022 - ui.adsabs.harvard.edu
Open-world instance segmentation is the task of grouping pixels into object instances
without any pre-determined taxonomy. This is challenging, as state-of-the-art methods rely …