CODEWITHSUNDEEP
tensorflowcomputer-visiondeep-learning
AashishKSahu
AashishKSahu

Reputation: 369

wrong output size after conv2d function

the image size is [m,32,32,3] (m = no. of training examples)

the filter size is [3,3,3,10]

stride = 1

padding = None

if I convolve this using tensorflow.nn.conv2d then the output shape should be this, according to the formula

out ={ ( 32 - 3 + 2*(0) ) / 1 }+ 1 = 30

so the output size should be [m, 30, 30, 10] but the output shape i am getting is [m, 32, 32, 10]

why is this happening?

# convolution layer 1
c1 = tf.nn.conv2d(x_train, w1, strides = [1,1,1,1], padding = 'SAME')
print('c1 size: ', c1.shape)
# activation function for c1: relu
r1 = tf.nn.relu(c1)
# maxpooling
p1 = tf.nn.max_pool(r1, ksize = [1,2,2,1], strides = [1,2,2,1], padding = 'SAME')

Upvotes: 1

Views: 583

Answers (1)

Vikash Singh
Vikash Singh

Reputation: 14001

padding = "SAME" means:

input       = [1, 2, 3, 4, 5, 6, 7, 8]
filter size = [1, 3]
stride      = [2]

so input to filter will be [[1, 2, 3], [3, 4, 5], [5, 6, 7], [7, 8, 0]]

padding = "VALID" means:

input       = [1, 2, 3, 4, 5, 6, 7, 8]
filter size = [1, 3]
stride      = [2]

so input to filter will be [[1, 2, 3], [3, 4, 5], [5, 6, 7]]

Last pixel got dropped in this case.

So padding "VALID" will give you the output you expect.

Upvotes: 3

Related Questions