prepare_mano.py

from config import *
from kinematics import *
import numpy as np
from utils import *


def prepare_mano():
  """
  Use this function to convert a mano model (from MANO-Hand Project) to the hand
  model we want to use in the project.
  """
  with open(OFFICIAL_MANO_PATH, 'rb') as f:
    data = pickle.load(f, encoding='latin1')

  output = {}
  output['verts'] = np.array(data['v_template'])
  output['faces'] = np.array(data['f'])
  output['mesh_basis'] = np.transpose(data['shapedirs'], (2, 0, 1))

  j_regressor = np.zeros([21, 778])
  j_regressor[:16] = data['J_regressor'].toarray()
  for k, v in MANOHandJoints.mesh_mapping.items():
    j_regressor[k, v] = 1
  output['j_regressor'] = j_regressor
  output['joints'] = np.matmul(output['j_regressor'], output['verts'])

  raw_weights = data['weights']
  weights = [None] * 21
  weights[0] = raw_weights[:, 0]
  for j in 'IMLRT':
    weights[MANOHandJoints.labels.index(j + '0')] = np.zeros(778)
    for k in [1, 2, 3]:
      src_idx = MANOHandJoints.labels.index(j + str(k - 1))
      tar_idx = MANOHandJoints.labels.index(j + str(k))
      weights[tar_idx] = raw_weights[:, src_idx]
  output['weights'] = np.expand_dims(np.stack(weights, -1), -1)
  with open(HAND_MESH_MODEL_PATH, 'wb') as f:
    pickle.dump(output, f)


if __name__ == '__main__':
  prepare_mano()