def _run_icp(self, cloud_source, ego_motion, cloud_target):
transform, residual = icp(cloud_source, ego_motion, cloud_target)
logging.info('Running ICP:')
logging.info('ego_motion: %s\n%s', ego_motion, ego_motion.eval())
logging.info('transform: %s\n%s', transform, transform.eval())
logging.info('residual: %s\n%s', residual,
residual[0, :PRINT_CAP, :].eval())
return transform, residual
def _run_icp(self, cloud_source, ego_motion, cloud_target):
transform, residual = icp(cloud_source, ego_motion, cloud_target)
logging.info('Running ICP:')
logging.info('ego_motion: %s\n%s', ego_motion, ego_motion.eval())
logging.info('transform: %s\n%s', transform, transform.eval())
logging.info('residual: %s\n%s', residual,
residual[0, :PRINT_CAP, :].eval())
return transform, residual
def inference(source, target):
"""Builds model."""
ego_motion = tf.Variable(tf.zeros([6]), name='ego_motion')
res_height = tf.Variable(tf.fill([1], 0.0), name='res_height')
tf.summary.scalar('tx', ego_motion[0])
tf.summary.scalar('ty', ego_motion[1])
tf.summary.scalar('tz', ego_motion[2])
tf.summary.scalar('rx', ego_motion[3])
tf.summary.scalar('ry', ego_motion[4])
tf.summary.scalar('rz', ego_motion[5])
tf.summary.scalar('res_height', res_height[0])
dist_to_center = tf.norm((source - RES_CENTER)[:, :, :2], axis=2,
keep_dims=True)
res = tf.maximum(RES_RADIUS - dist_to_center, 0.0) / RES_RADIUS
res *= res_height
res = tf.concat([tf.zeros_like(res), tf.zeros_like(res), res], axis=2)
shifted_source = source + res
ego_motion = tf.stack([ego_motion] * FLAGS.batch_size)
transform, residual = icp(shifted_source, ego_motion, target)
return transform, residual
def inference(source, target):
"""Builds model."""
ego_motion = tf.Variable(tf.zeros([6]), name='ego_motion')
res_height = tf.Variable(tf.fill([1], 0.0), name='res_height')
tf.summary.scalar('tx', ego_motion[0])
tf.summary.scalar('ty', ego_motion[1])
tf.summary.scalar('tz', ego_motion[2])
tf.summary.scalar('rx', ego_motion[3])
tf.summary.scalar('ry', ego_motion[4])
tf.summary.scalar('rz', ego_motion[5])
tf.summary.scalar('res_height', res_height[0])
dist_to_center = tf.norm((source - RES_CENTER)[:, :, :2], axis=2,
keep_dims=True)
res = tf.maximum(RES_RADIUS - dist_to_center, 0.0) / RES_RADIUS
res *= res_height
res = tf.concat([tf.zeros_like(res), tf.zeros_like(res), res], axis=2)
shifted_source = source + res
ego_motion = tf.stack([ego_motion] * FLAGS.batch_size)
transform, residual = icp(shifted_source, ego_motion, target)
return transform, residual
