CODEWITHSUNDEEP
tensorflow
save_ole
save_ole

Reputation: 310

Calculate sum inside a tensor

I am designing a neural network with Tensorflow. I have a tensor G with shape: [batch_size, C, 1 + x + y, feature_size], I want to calculate the sum of x and y vectors inside G, so that the new G has shape: [batch_size, C, 3, feature_size]. So which operation in Tensorflow should I use?

Thank you for your help.

Upvotes: 0

Views: 99

Answers (1)

Zvika
Zvika

Reputation: 1280

Try the following:

inp = tf.placeholder(tf.float32, shape=(batch_size, C, 1+x+y, feature_size))
s1 = inp[:, :, 0:1, :]
sx = tf.reduce_sum(inp[:, :, 1:x+1, :], axis=2, keepdims=True)
sy = tf.reduce_sum(inp[:, :, -y:, :], axis=2, keepdums=True)
result = tf.concat((s1, sx, sy), axis=2)

Upvotes: 1

Related Questions