This repository contains an implementation for performing 3D animal (quadruped) reconstruction from a monocular image or video. The system adapts the pose (limb positions) and shape (animal type/height/weight) parameters for the SMAL deformable quadruped model, as well as camera parameters until the projected SMAL model aligns with 2D keypoints and silhouette segmentations extracted from the input frame(s).