Work

BSSA-net: category-level 6D object pose estimation with bi-scale and semantic-aligned feature interaction