Chirality Nets for Human Pose Regression

The proposed layers lead toamore data efficient representation and areduction in computation by exploiting symmetry. We evaluate chirality nets on the task ofhuman poseregression, which naturally exploits theleft/right mirroring ofthe human body.