CVPR2021

Heterogeneous Grid Convolution for Adaptive, Efficient, and Controllable Computation

Ryuhei Hamaguchi, Yasutaka Furukawa, Masaki Onishi, Ken Sakurada

摘要

Figure 1. Heterogeneous grid convolution exploits the heterogeneity in the image to enable adaptive, efficient, and controllable computation for a range of image understanding tasks such as semantic segmentation, road extraction, and salient object detection from left to right.