Papers by Ziling Huang
Referring Image Segmentation via Joint Mask Contextual Embedding Learning and Progressive Alignment Network (2023.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for referring image segmentation may encounter limitations in maintaining focus on relevant information during specific stages and rectifying errors propagated from early stages. |
| Approach: | They propose a network that integrates a Learnable Contextual Embedding module and a Progressive Alignment Network to enhance the cascade framework. |
| Outcome: | The proposed network achieves state-of-the-art results on three commonly used benchmarks. |