Dynamic graph message passing networks
Modelling long-range dependencies is critical for scene understanding tasks in computer vision. Although CNNs have excelled in many vision tasks, they are still limited in capturing long-range structured relationships as they typically consist of layers of local kernels. A fully-connected graph is b...
| المؤلفون الرئيسيون: | , , , |
|---|---|
| التنسيق: | Conference item |
| اللغة: | English |
| منشور في: |
IEEE
2020
|