def batch_tps_rpm_bij(
src_ctx,
tgt_ctx,
T_init=1e-1,
T_final=5e-3,
outlierfrac=1e-2,
outlierprior=1e-1,
outliercutoff=1e-2,
em_iter=EM_ITER_CHEAP,
component_cost=False,
):
"""
computes tps rpm for the clouds in src and tgt in batch
TODO: Fill out comment cleanly
"""
##TODO: add check to ensure that src_ctx and tgt_ctx are formatted properly
n_iter = len(src_ctx.bend_coefs)
T_vals = loglinspace(T_init, T_final, n_iter)
src_ctx.reset_tps_params()
tgt_ctx.reset_tps_params()
for i, b in enumerate(src_ctx.bend_coefs):
T = T_vals[i]
for _ in range(em_iter):
src_ctx.transform_points()
tgt_ctx.transform_points()
src_ctx.get_target_points(tgt_ctx, outlierprior, outlierfrac, outliercutoff, T)
src_ctx.update_transform(b)
# check_update(src_ctx, b)
tgt_ctx.update_transform(b)
return src_ctx.bidir_tps_cost(tgt_ctx, return_components=component_cost)
def __init__(self, bend_coefs=None):
if bend_coefs is None:
lambda_init, lambda_final = DEFAULT_LAMBDA
bend_coefs = np.around(loglinspace(lambda_init, lambda_final, N_ITER_CHEAP), BEND_COEF_DIGITS)
self.bend_coefs = bend_coefs
self.ptrs_valid = False
self.N = 0
self.tps_params = []
self.tps_param_ptrs = None
self.trans_d = []
self.trans_d_ptrs = None
self.lin_dd = []
self.lin_dd_ptrs = None
self.w_nd = []
self.w_nd_ptrs = None
"""
TPS PARAM FORMAT
[ np.zeros(DATA_DIM) ] [ trans_d ] [1 x d]
[ np.eye(DATA_DIM) ] = [ lin_dd ] = [d x d]
[np.zeros((np.zeros, DATA_DIM))] [ w_nd ] [n x d]
"""
self.default_tps_params = gpuarray.zeros((DATA_DIM + 1 + MAX_CLD_SIZE, DATA_DIM), np.float32)
self.default_tps_params[1 : DATA_DIM + 1, :].set(np.eye(DATA_DIM, dtype=np.float32))
self.proj_mats = dict([(b, []) for b in bend_coefs])
self.proj_mat_ptrs = dict([(b, None) for b in bend_coefs])
self.offset_mats = dict([(b, []) for b in bend_coefs])
self.offset_mat_ptrs = dict([(b, None) for b in bend_coefs])
self.pts = []
self.pt_ptrs = None
self.kernels = []
self.kernel_ptrs = None
self.pts_w = []
self.pt_w_ptrs = None
self.pts_t = []
self.pt_t_ptrs = None
self.dims = []
self.dims_gpu = None
self.scale_params = []
self.warp_err = None
self.bend_res = []
self.bend_res_ptrs = None
self.corr_cm = []
self.corr_cm_ptrs = None
self.corr_rm = []
self.corr_rm_ptrs = None
self.r_coefs = []
self.r_coef_ptrs = None
self.c_coefs_rn = []
self.c_coef_rn_ptrs = None
self.c_coefs_cn = []
self.c_coef_cn_ptrs = None
self.seg_names = []
self.names2inds = {}
def test_batch_tps_rpm_bij(
src_ctx,
tgt_ctx,
T_init=1e-1,
T_final=5e-3,
outlierfrac=1e-2,
outlierprior=1e-1,
outliercutoff=0.5,
em_iter=EM_ITER_CHEAP,
test_ind=0,
):
from lfd.tpsopt.transformations import ThinPlateSpline, set_ThinPlateSpline
n_iter = len(src_ctx.bend_coefs)
T_vals = loglinspace(T_init, T_final, n_iter)
x_nd = src_ctx.pts[test_ind].get()[: src_ctx.dims[test_ind]]
y_md = tgt_ctx.pts[0].get()[: tgt_ctx.dims[0]]
(n, d) = x_nd.shape
(m, _) = y_md.shape
f = ThinPlateSpline(d)
g = ThinPlateSpline(d)
src_ctx.reset_tps_params()
tgt_ctx.reset_tps_params()
for i, b in enumerate(src_ctx.bend_coefs):
T = T_vals[i]
for _ in range(em_iter):
src_ctx.transform_points()
tgt_ctx.transform_points()
xwarped_nd = f.transform_points(x_nd)
ywarped_md = g.transform_points(y_md)
gpu_xw = src_ctx.pts_w[test_ind].get()[:n, :]
gpu_yw = tgt_ctx.pts_w[test_ind].get()[:m, :]
assert np.allclose(xwarped_nd, gpu_xw, atol=1e-5)
assert np.allclose(ywarped_md, gpu_yw, atol=1e-5)
xwarped_nd = gpu_xw
ywarped_md = gpu_yw
src_ctx.get_target_points(tgt_ctx, outlierprior, outlierfrac, outliercutoff, T)
fwddist_nm = ssd.cdist(xwarped_nd, y_md, "euclidean")
invdist_nm = ssd.cdist(x_nd, ywarped_md, "euclidean")
prob_nm = outlierprior * np.ones((n + 1, m + 1), np.float32)
prob_nm[:n, :m] = np.exp(-(fwddist_nm + invdist_nm) / float(2 * T))
prob_nm[n, m] = outlierfrac * np.sqrt(n * m)
gpu_corr = src_ctx.corr_rm[test_ind].get()
gpu_corr = gpu_corr.flatten()
gpu_corr = gpu_corr[: (n + 1) * (m + 1)].reshape(n + 1, m + 1).astype(np.float32)
assert np.allclose(prob_nm[:n, :m], gpu_corr[:n, :m], atol=1e-5)
save_prob_nm = np.array(prob_nm)
save_gpu_corr = np.array(gpu_corr)
prob_nm[:n, :m] = gpu_corr[:n, :m]
r_coefs = np.ones(n + 1, np.float32)
c_coefs = np.ones(m + 1, np.float32)
a_N = np.ones((n + 1), dtype=np.float32)
a_N[n] = m * outlierfrac
b_M = np.ones((m + 1), dtype=np.float32)
b_M[m] = n * outlierfrac
for norm_iter_i in range(DEFAULT_NORM_ITERS):
r_coefs = a_N / prob_nm.dot(c_coefs)
rn_c_coefs = c_coefs
c_coefs = b_M / r_coefs.dot(prob_nm)
gpu_r_coefs = src_ctx.r_coefs[test_ind].get()[: n + 1].reshape(n + 1)
gpu_c_coefs_cn = src_ctx.c_coefs_cn[test_ind].get()[: m + 1].reshape(m + 1)
gpu_c_coefs_rn = src_ctx.c_coefs_rn[test_ind].get()[: m + 1].reshape(m + 1)
r_diff = np.abs(r_coefs - gpu_r_coefs)
rn_diff = np.abs(rn_c_coefs - gpu_c_coefs_rn)
cn_diff = np.abs(c_coefs - gpu_c_coefs_cn)
assert np.allclose(r_coefs, gpu_r_coefs, atol=1e-5)
assert np.allclose(c_coefs, gpu_c_coefs_cn, atol=1e-5)
assert np.allclose(rn_c_coefs, gpu_c_coefs_rn, atol=1e-5)
prob_nm = prob_nm[:n, :m]
prob_nm *= gpu_r_coefs[:n, None]
rn_p_nm = prob_nm * gpu_c_coefs_rn[None, :m]
cn_p_nm = prob_nm * gpu_c_coefs_cn[None, :m]
wt_n = rn_p_nm.sum(axis=1)
gpu_corr_cm = src_ctx.corr_cm[test_ind].get().flatten()[: (n + 1) * (m + 1)]
gpu_corr_cm = gpu_corr_cm.reshape(m + 1, n + 1) ## b/c it is column major
assert np.allclose(wt_n, gpu_corr_cm[m, :n], atol=1e-4)
inlier = wt_n > outliercutoff
xtarg_nd = np.empty((n, DATA_DIM), np.float32)
xtarg_nd[inlier, :] = rn_p_nm.dot(y_md)[inlier, :]
xtarg_nd[~inlier, :] = xwarped_nd[~inlier, :]
wt_m = cn_p_nm.sum(axis=0)
assert np.allclose(wt_m, gpu_corr[n, :m], atol=1e-4)
inlier = wt_m > outliercutoff
ytarg_md = np.empty((m, DATA_DIM), np.float32)
ytarg_md[inlier, :] = cn_p_nm.T.dot(x_nd)[inlier, :]
ytarg_md[~inlier, :] = ywarped_md[~inlier, :]
xt_gpu = src_ctx.pts_t[test_ind].get()[:n, :]
yt_gpu = tgt_ctx.pts_t[test_ind].get()[:m, :]
assert np.allclose(xtarg_nd, xt_gpu, atol=1e-4)
assert np.allclose(ytarg_md, yt_gpu, atol=1e-4)
src_ctx.update_transform(b)
tgt_ctx.update_transform(b)
f_p_mat = src_ctx.proj_mats[b][test_ind].get()[: n + d + 1, :n]
f_o_mat = src_ctx.offset_mats[b][test_ind].get()[: n + d + 1]
b_p_mat = tgt_ctx.proj_mats[b][0].get()[: m + d + 1, :m]
b_o_mat = tgt_ctx.offset_mats[b][0].get()[: m + d + 1]
f_params = f_p_mat.dot(xtarg_nd) + f_o_mat
g_params = b_p_mat.dot(ytarg_md) + b_o_mat
gpu_fparams = src_ctx.tps_params[test_ind].get()[: n + d + 1]
gpu_gparams = tgt_ctx.tps_params[test_ind].get()[: m + d + 1]
assert np.allclose(f_params, gpu_fparams, atol=1e-4)
assert np.allclose(g_params, gpu_gparams, atol=1e-4)
set_ThinPlateSpline(f, x_nd, gpu_fparams)
set_ThinPlateSpline(g, y_md, gpu_gparams)
f._cost = tps.tps_cost(f.lin_ag, f.trans_g, f.w_ng, f.x_na, xtarg_nd, 1)
g._cost = tps.tps_cost(g.lin_ag, g.trans_g, g.w_ng, g.x_na, ytarg_md, 1)
gpu_cost = src_ctx.bidir_tps_cost(tgt_ctx)
cpu_cost = f._cost + g._cost
assert np.isclose(gpu_cost[test_ind], cpu_cost, atol=1e-4)
def main():
scikits.cuda.linalg.init()
args = parse_arguments()
f = h5py.File(args.datafile, 'r+')
bend_coefs = np.around(
loglinspace(args.bend_coef_init, args.bend_coef_final, args.n_iter),
BEND_COEF_DIGITS)
for seg_name, seg_info in f.iteritems():
if 'inv' in seg_info:
if args.replace:
del seg_info['inv']
inv_group = seg_info.create_group('inv')
else:
inv_group = seg_info['inv']
else:
inv_group = seg_info.create_group('inv')
ds_key = 'DS_SIZE_{}'.format(DS_SIZE)
if ds_key in inv_group:
scaled_x_na = inv_group[ds_key]['scaled_cloud_xyz'][:]
K_nn = inv_group[ds_key]['scaled_K_nn'][:]
else:
ds_g = inv_group.create_group(ds_key)
x_na = downsample_cloud(seg_info[args.cloud_name][:, :3])
scaled_x_na, scale_params = unit_boxify(x_na)
K_nn = tps_kernel_matrix(scaled_x_na)
if args.fill_traj:
r_traj = seg_info['r_gripper_tool_frame']['hmat'][:, :3, 3]
l_traj = seg_info['l_gripper_tool_frame']['hmat'][:, :3, 3]
scaled_r_traj = r_traj * scale_params[0] + scale_params[1]
scaled_l_traj = l_traj * scale_params[0] + scale_params[1]
scaled_r_traj_K = tps_kernel_matrix2(scaled_r_traj,
scaled_x_na)
scaled_l_traj_K = tps_kernel_matrix2(scaled_l_traj,
scaled_x_na)
ds_g['scaled_r_traj'] = scaled_r_traj
ds_g['scaled_l_traj'] = scaled_l_traj
ds_g['scaled_r_traj_K'] = scaled_r_traj_K
ds_g['scaled_l_traj_K'] = scaled_l_traj_K
# Precompute l,r closing indices
lr2finger_traj = {}
for lr in 'lr':
arm_name = {"l": "leftarm", "r": "rightarm"}[lr]
lr2finger_traj[
lr] = gripper_joint2gripper_l_finger_joint_values(
np.asarray(seg_info['%s_gripper_joint' %
lr]))[:, None]
opening_inds, closing_inds = get_opening_closing_inds(
lr2finger_traj[lr])
if '%s_closing_inds' % lr in seg_info:
del seg_info['%s_closing_inds' % lr]
if not closing_inds:
closing_inds = False
seg_info['%s_closing_inds' % lr] = closing_inds
#
ds_g['cloud_xyz'] = x_na
ds_g['scaled_cloud_xyz'] = scaled_x_na
ds_g['scaling'] = scale_params[0]
ds_g['scaled_translation'] = scale_params[1]
ds_g['scaled_K_nn'] = K_nn
for bend_coef in bend_coefs:
if str(bend_coef) in inv_group:
continue
bend_coef_g = inv_group.create_group(str(bend_coef))
_, res = get_sol_params(scaled_x_na, K_nn, bend_coef)
for k, v in res.iteritems():
bend_coef_g[k] = v
if args.verbose:
sys.stdout.write(
'\rprecomputed tps solver for segment {}'.format(seg_name))
sys.stdout.flush()
print ""
if args.test:
atol = 1e-7
print 'Running batch get sol params test with atol = ', atol
test_batch_get_sol_params(f, [bend_coef], atol=atol)
print 'batch sol params test succeeded'
print ""
bend_coefs = np.around(
loglinspace(args.exact_bend_coef_init, args.exact_bend_coef_final,
args.exact_n_iter), BEND_COEF_DIGITS)
for seg_name, seg_info in f.iteritems():
if 'solver' in seg_info:
if args.replace:
del seg_info['solver']
solver_g = seg_info.create_group('solver')
else:
solver_g = seg_info['solver']
else:
solver_g = seg_info.create_group('solver')
x_nd = seg_info['inv'][ds_key]['scaled_cloud_xyz'][:]
K_nn = seg_info['inv'][ds_key]['scaled_K_nn'][:]
N, QN, NON, NR = get_exact_solver(x_nd, K_nn, bend_coefs)
solver_g['N'] = N
solver_g['QN'] = QN
solver_g['NR'] = NR
solver_g['x_nd'] = x_nd
solver_g['K_nn'] = K_nn
NON_g = solver_g.create_group('NON')
for b in bend_coefs:
NON_g[str(b)] = NON[b]
if args.verbose:
sys.stdout.write(
'\rprecomputed exact tps solver for segment {}'.format(
seg_name))
sys.stdout.flush()
print ""
f.close()
def main():
scikits.cuda.linalg.init()
args = parse_arguments()
f = h5py.File(args.datafile, 'r+')
bend_coefs = np.around(loglinspace(args.bend_coef_init, args.bend_coef_final, args.n_iter),
BEND_COEF_DIGITS)
for seg_name, seg_info in f.iteritems():
if 'inv' in seg_info:
if args.replace:
del seg_info['inv']
inv_group = seg_info.create_group('inv')
else:
inv_group = seg_info['inv']
else:
inv_group = seg_info.create_group('inv')
ds_key = 'DS_SIZE_{}'.format(DS_SIZE)
if ds_key in inv_group:
scaled_x_na = inv_group[ds_key]['scaled_cloud_xyz'][:]
K_nn = inv_group[ds_key]['scaled_K_nn'][:]
else:
ds_g = inv_group.create_group(ds_key)
x_na = downsample_cloud(seg_info[args.cloud_name][:, :3])
scaled_x_na, scale_params = unit_boxify(x_na)
K_nn = tps_kernel_matrix(scaled_x_na)
if args.fill_traj:
r_traj = seg_info['r_gripper_tool_frame']['hmat'][:, :3, 3]
l_traj = seg_info['l_gripper_tool_frame']['hmat'][:, :3, 3]
scaled_r_traj = r_traj * scale_params[0] + scale_params[1]
scaled_l_traj = l_traj * scale_params[0] + scale_params[1]
scaled_r_traj_K = tps_kernel_matrix2(scaled_r_traj, scaled_x_na)
scaled_l_traj_K = tps_kernel_matrix2(scaled_l_traj, scaled_x_na)
ds_g['scaled_r_traj'] = scaled_r_traj
ds_g['scaled_l_traj'] = scaled_l_traj
ds_g['scaled_r_traj_K'] = scaled_r_traj_K
ds_g['scaled_l_traj_K'] = scaled_l_traj_K
# Precompute l,r closing indices
lr2finger_traj = {}
for lr in 'lr':
arm_name = {"l":"leftarm", "r":"rightarm"}[lr]
lr2finger_traj[lr] = gripper_joint2gripper_l_finger_joint_values(np.asarray(seg_info['%s_gripper_joint'%lr]))[:,None]
opening_inds, closing_inds = get_opening_closing_inds(lr2finger_traj[lr])
if '%s_closing_inds'%lr in seg_info:
del seg_info['%s_closing_inds'%lr]
if not closing_inds:
closing_inds = False
seg_info['%s_closing_inds'%lr] = closing_inds
#
ds_g['cloud_xyz'] = x_na
ds_g['scaled_cloud_xyz'] = scaled_x_na
ds_g['scaling'] = scale_params[0]
ds_g['scaled_translation'] = scale_params[1]
ds_g['scaled_K_nn'] = K_nn
for bend_coef in bend_coefs:
if str(bend_coef) in inv_group:
continue
bend_coef_g = inv_group.create_group(str(bend_coef))
_, res = get_sol_params(scaled_x_na, K_nn, bend_coef)
for k, v in res.iteritems():
bend_coef_g[k] = v
if args.verbose:
sys.stdout.write('\rprecomputed tps solver for segment {}'.format(seg_name))
sys.stdout.flush()
print ""
if args.test:
atol = 1e-7
print 'Running batch get sol params test with atol = ', atol
test_batch_get_sol_params(f, [bend_coef], atol=atol)
print 'batch sol params test succeeded'
print ""
bend_coefs = np.around(loglinspace(args.exact_bend_coef_init, args.exact_bend_coef_final,
args.exact_n_iter),
BEND_COEF_DIGITS)
for seg_name, seg_info in f.iteritems():
if 'solver' in seg_info:
if args.replace:
del seg_info['solver']
solver_g = seg_info.create_group('solver')
else:
solver_g = seg_info['solver']
else:
solver_g = seg_info.create_group('solver')
x_nd = seg_info['inv'][ds_key]['scaled_cloud_xyz'][:]
K_nn = seg_info['inv'][ds_key]['scaled_K_nn'][:]
N, QN, NON, NR = get_exact_solver(x_nd, K_nn, bend_coefs)
solver_g['N'] = N
solver_g['QN'] = QN
solver_g['NR'] = NR
solver_g['x_nd'] = x_nd
solver_g['K_nn'] = K_nn
NON_g = solver_g.create_group('NON')
for b in bend_coefs:
NON_g[str(b)] = NON[b]
if args.verbose:
sys.stdout.write('\rprecomputed exact tps solver for segment {}'.format(seg_name))
sys.stdout.flush()
print ""
f.close()
