def get_mpi_affective_features(data):
# 0: root, 1: chest, 2: chest2, 3: chest3,
# 4: chest4, 5: neck, 6: head,
# 7: left_collar, 8: left_shoulder, 9: left_elbow, 10: left_wrist,
# 11: right_collar, 12: right_shoulder, 13: right_elbow, 14: right_wrist,
# 15: left_hip, 16: left_knee, 17: left_heel, 18: left_toe,
# 19: right_hip, 20: right_knee, 21: right_heel, 22: right_toe
affs_dim = 15
if type(data).__module__ != np.__name__:
data_np = data.detach().cpu().numpy()
else:
data_np = data
_0, _1, _2, _3, _4, _5, _6, _7, _8, _9, _10, _11, _12,\
_13, _14, _15, _16, _17, _18, _19, _20, _21, _22 = MocapDataset.get_mpi_joints(data_np)
affective_features = np.zeros(data_np.shape[:-2] + (affs_dim,))
fidx = 0
affective_features[..., fidx] = common.angle_between_points(_0, _2, _6) / np.pi
fidx += 1
affective_features[..., fidx] = common.angle_between_points(_9, _4, _6) / np.pi
fidx += 1
affective_features[..., fidx] = common.angle_between_points(_13, _4, _6) / np.pi
fidx += 1
affective_features[..., fidx] = common.angle_between_points(_0, _10, _4) / np.pi
fidx += 1
affective_features[..., fidx] = common.angle_between_points(_0, _14, _4) / np.pi
fidx += 1
affective_features[..., fidx] = common.angle_between_points(_8, _9, _10) / np.pi
fidx += 1
affective_features[..., fidx] = common.angle_between_points(_12, _13, _14) / np.pi
fidx += 1
affective_features[..., fidx] = common.dist_between(_10, _0) / common.dist_between(_10, _4)
fidx += 1
affective_features[..., fidx] = common.dist_between(_10, _0) / common.dist_between(_10, _6)
fidx += 1
affective_features[..., fidx] = common.dist_between(_14, _0) / common.dist_between(_14, _4)
fidx += 1
affective_features[..., fidx] = common.dist_between(_14, _0) / common.dist_between(_14, _6)
fidx += 1
affective_features[..., fidx] = common.dist_between(_10, _14) / common.dist_between(_0, _6)
fidx += 1
affective_features[..., fidx] = common.area_of_triangle(_10, _0, _14) / common.area_of_triangle(_10, _4, _14)
fidx += 1
affective_features[..., fidx] = common.area_of_triangle(_6, _10, _0) / common.area_of_triangle(_6, _14, _0)
fidx += 1
affective_features[..., fidx] = common.area_of_triangle(_4, _10, _0) / common.area_of_triangle(_4, _14, _0)
fidx += 1
affective_features = np.nan_to_num(affective_features)
if type(data).__module__ != np.__name__:
affective_features = torch.from_numpy(affective_features).cuda()
return affective_features
def get_affective_features(data):
# 0: root, 1: left_hip, 2: left_knee, 3: left_heel, 4: left_toe,
# 5: right_hip, 6: right_knee, 7: right_heel, 8: right_toe,
# 9: lower_back, 10: spine, 11: neck, 12: head,
# 13: left_shoulder, 14: left_elbow, 15: left_hand, 16: left_hand_index,
# 17: right_shoulder, 18: right_elbow, 19: right_hand, 20: right_hand_index
affs_dim = 18
_0, _1, _2, _3, _4, _5, _6, _7, _8, _9, _10,\
_11, _12, _13, _14, _15, _16, _17, _18, _19, _20 = MocapDataset.get_joints(data)
affective_features = np.zeros(data.shape[:-2] + (affs_dim, ))
fidx = 0
affective_features[...,
fidx] = common.angle_between_points(_17, _9, _13)
fidx += 1
affective_features[...,
fidx] = common.angle_between_points(_19, _9, _15)
fidx += 1
affective_features[..., fidx] = common.dist_between(
_16, _11) / common.dist_between(_16, _0)
fidx += 1
affective_features[..., fidx] = common.dist_between(
_20, _11) / common.dist_between(_20, _0)
fidx += 1
affective_features[..., fidx] = common.dist_between(
_16, _20) / common.dist_between(_11, _0)
fidx += 1
affective_features[..., fidx] = \
common.area_of_triangle(_17, _9, _13) / common.area_of_triangle(_17, _0, _13)
fidx += 1
affective_features[..., fidx] = \
common.area_of_triangle(_19, _9, _15) / common.area_of_triangle(_19, _0, _15)
fidx += 1
affective_features[...,
fidx] = common.angle_between_points(_13, _14, _15)
fidx += 1
affective_features[...,
fidx] = common.angle_between_points(_17, _18, _19)
fidx += 1
affective_features[...,
fidx] = common.angle_between_points(_12, _11, _13)
fidx += 1
affective_features[...,
fidx] = common.angle_between_points(_12, _11, _17)
fidx += 1
affective_features[...,
fidx] = common.angle_between_points(_12, _0, _2)
fidx += 1
affective_features[...,
fidx] = common.angle_between_points(_12, _0, _6)
fidx += 1
affective_features[..., fidx] = common.dist_between(
_4, _8) / common.dist_between(_11, _0)
fidx += 1
affective_features[..., fidx] = common.angle_between_points(_8, _0, _4)
fidx += 1
affective_features[..., fidx] = common.angle_between_points(_1, _2, _3)
fidx += 1
affective_features[..., fidx] = common.angle_between_points(_5, _6, _7)
fidx += 1
affective_features[..., fidx] = \
common.area_of_triangle(_20, _11, _16) / common.area_of_triangle(_8, _0, _4)
fidx += 1
return np.nan_to_num(affective_features)
def get_affective_features(data):
# 0: hips,
# 1: left_hip_joint, 2: left_up_leg, 3: left_leg, 4: left_foot, 5: left_toe_base,
# 6: right_hip_joint, 7: right_up_leg, 8: right_leg, 9: right_foot, 10: right_toe_base,
# 11: lower_back, 12: spine, 13: spine1,
# 14: neck, 15: neck1, 16: head,
# 17: left_shoulder, 18: left_arm, 19: left_forearm, 20: left_hand,
# 21: left_finger_base, 22: left_hand_index1, 23: left_thumb,
# 24: right_shoulder, 25: right_arm, 26: right_forearm, 27: right_hand,
# 28: right_finger_base, 29: right_hand_index1, 30: right_thumb,
affs_dim = 18
_0, _1, _2, _3, _4, _5, _6, _7, _8, _9, _10, _11, _12, _13, _14, _15,\
_16, _17, _18, _19, _20, _21, _22, _23, _24, _25, _26, _27, _28, _29, _30 = MocapDataset.get_joints(data)
affective_features = np.zeros(data.shape[:-2] + (affs_dim, ))
fidx = 0
affective_features[...,
fidx] = common.angle_between_points(_25, _12, _18)
fidx += 1
affective_features[...,
fidx] = common.angle_between_points(_27, _12, _20)
fidx += 1
affective_features[..., fidx] = common.dist_between(
_22, _15) / common.dist_between(_22, _0)
fidx += 1
affective_features[..., fidx] = common.dist_between(
_29, _15) / common.dist_between(_29, _0)
fidx += 1
affective_features[..., fidx] = common.dist_between(
_22, _29) / common.dist_between(_15, _0)
fidx += 1
affective_features[..., fidx] = \
common.area_of_triangle(_25, _12, _18) / common.area_of_triangle(_25, _0, _18)
fidx += 1
affective_features[..., fidx] = \
common.area_of_triangle(_27, _12, _20) / common.area_of_triangle(_27, _0, _20)
fidx += 1
affective_features[...,
fidx] = common.angle_between_points(_18, _19, _20)
fidx += 1
affective_features[...,
fidx] = common.angle_between_points(_25, _26, _27)
fidx += 1
affective_features[...,
fidx] = common.angle_between_points(_16, _15, _18)
fidx += 1
affective_features[...,
fidx] = common.angle_between_points(_16, _15, _25)
fidx += 1
affective_features[...,
fidx] = common.angle_between_points(_16, _0, _3)
fidx += 1
affective_features[...,
fidx] = common.angle_between_points(_16, _0, _8)
fidx += 1
affective_features[..., fidx] = common.dist_between(
_5, _10) / common.dist_between(_15, _0)
fidx += 1
affective_features[...,
fidx] = common.angle_between_points(_10, _0, _5)
fidx += 1
affective_features[..., fidx] = common.angle_between_points(_2, _3, _4)
fidx += 1
affective_features[..., fidx] = common.angle_between_points(_7, _8, _9)
fidx += 1
affective_features[..., fidx] = \
common.area_of_triangle(_29, _15, _22) / common.area_of_triangle(_10, _0, _5)
fidx += 1
return np.nan_to_num(affective_features)