Regarding the use of tf.tile? - python-3.x

Question

I am trying to get the numpy code into tensorflow. But when I run the session I am getting the below error.
Here is the code:
def cal_anchors():
# Output:
# anchors: (w, l, 2, 7) x y z h w l r
x = np.linspace(0,48,240) #48/.2
y = np.linspace(20,-20,200) #40/.2
cx, cy = np.meshgrid(x, y)
# all is (w, l, 2)
cx = np.tile(cx[..., np.newaxis], 2)
cy = np.tile(cy[..., np.newaxis], 2)
cz = np.ones_like(cx) * (-1.465)
w = np.ones_like(cx) * 0.6
l = np.ones_like(cx) * 0.8
h = np.ones_like(cx) * 1.73
r = np.ones_like(cx) #(200,240,2)
print(r.shape)
r[..., 0] = 0 # 0 #(200,240,1)
r[..., 1] = 90 / 180 * np.pi # 90
# 7*(w,l,2) -> (w, l, 2, 7)
anchors = np.stack([cx, cy, cz, h, w, l, r], axis=-1)
return anchors
print(cal_anchors().shape) #(200,240,2,7)
#tensorflow code
def anchors():
x = tf.linspace(0.0,48.0,240)
y = tf.linspace(-20.0,20.0,200)
cx,cy = tf.meshgrid(x,y)
cx = tf.expand_dims(cx,2)
cy = tf.expand_dims(cy,2)
cx = tf.tile(cx,2)
cy = tf.tile(cy,2)
cz = tf.ones_like(cx)
w = tf.ones_like(cx)
l = tf.ones_like(cx)
h = tf.ones_like(cx)
r = ones_like(cx)
r[:,:,0] = 0
pi = tf.constant(3.141592)
r[:,:,1] = tf.math.divide(90,tf.math.multiply(180,pi))
anchros = tf.stack([cx,cy,cz,h,w,l,r],axis=-1,name="anchors")
sess = tf.Session()
sess.run(tf.global_variables_initializer())
result = sess.run(anchros)
print(result.shape)
anchors()
I am getting the below error:
File "anchros.py", line 39, in anchors
cx = tf.tile(cx,2)
File "/home/surendra/venv/tensorflow2/lib/python3.6/site-packages/tensorflow/python/ops/gen_array_ops.py", line 11450, in tile
_ops.raise_from_not_ok_status(e, name)
File "/home/surendra/venv/tensorflow2/lib/python3.6/site-packages/tensorflow/python/framework/ops.py", line 6862, in raise_from_not_ok_status
six.raise_from(core._status_to_exception(e.code, message), None)
File "", line 3, in raise_from
tensorflow.python.framework.errors_impl.InvalidArgumentError: Expected multiples to be 1-D, but got shape [] [Op:Tile]
Can anyone help me with this?
Thanks in advance

Answer 1

Couple of issues there, this should work:
def anchors():
x = tf.linspace(0.0, 48.0, 240)
y = tf.linspace(-20.0, 20.0, 200)
cx, cy = tf.meshgrid(x, y)
cx = tf.expand_dims(cx, 2)
cy = tf.expand_dims(cy, 2)
cx = tf.tile(cx, [1,1,2])
cy = tf.tile(cy, [1,1,2])
cz = tf.ones_like(cx)
w = tf.ones_like(cx)
l = tf.ones_like(cx)
h = tf.ones_like(cx)
r = tf.Variable(initial_value=tf.ones_like(cx))
r[:, :, 0].assign(tf.zeros(r.shape[:2]))
pi = tf.constant(3.141592)
r[:, :, 1].assign(tf.multiply(tf.math.divide(90., tf.math.multiply(180., pi)), tf.ones(r.shape[:2])))
anchros = tf.stack([cx, cy, cz, h, w, l, r], axis=-1, name="anchors")
sess = tf.Session()
sess.run(tf.global_variables_initializer())
result = sess.run(anchros)
print(result.shape)