:param num_traj: number of trajectories (int) :param env: environment (env class, str with env_name, or factory function) :par

Convert Euler Angles to Quaternions. See rotation.py for notes

A definition of relevant dimensions {all_b, nh, t, d}: image_size: Side length of input images (assumed square) all_b

:param model: pytorch model of form y_hat = f(x) (class) :param x: inputs to the model (tensor) :param y:

flip coords

first generate enough perturbed actions then do rollouts with generated actions set seed inside this function for multiprocessing

:param joints: [num_joints, 3] :param joints_vis: [num_joints, 3] :return: target, target_weight(1: visible, 0: invisible)

get predictions from score maps heatmaps: numpy.ndarray([batch_size, num_joints, height, width])

Convert Rotation Matrix to Euler Angles. See rotation.py for notes

Due to specifics of new mujoco rendering, the standard viewer cannot be used with this set-up. Instead we use this mujoco specific fu

Convert Quaternion to Euler Angles. See rotation.py for notes

Convert Quaternion to Euler Angles. See rotation.py for notes

Reset the robot degrees of freedom (qpos and qvel). Implement this in each subclass.

sample a batch of videos