no code implementations • 20 Oct 2021 • Noriaki Hirose, Kosuke Tahara
This framework is attractive for researchers because the depth and pose networks can be trained from just time sequence images without the need for the ground truth depth and poses.