Co-occurring keywords · papers mentioning both
Papers
Self-Mutual Distillation Learning for Continuous Sign Language Recognition
ICCV 2021
Specialize and Fuse: Pyramidal Output Representation for Semantic Segmentation
ICCV 2021