Reconstructing the 3D shape and pose of humans and animals is essential in many future applications, such as autonomous vehicle’s Forward Collision Avoidance Assist (FCA) algorithms. In this work, we leverage well designed low dimensional linear mesh models of human and animal, SMPL and SMAL, to jointly regress the 3D mesh for a single input RGB image. We show that our joint regression network can learn anatomical similarities among humans and other various animal species.