mvNCCompile successfully compiles my model, but I get wrong result of slice operation.
The input shape of my model is (1,30,20,4), the code is shown as follows:
image = tf.placeholder(tf.float32, shape=(1, 30, 20, 4),name='X')
image1 = tf.slice(image,[0,0,0,0],[1,30,20,2],name='s1')
image2 = tf.slice(image,[0,0,0,2],[1,30,20,2],name='s2')
The result image1 is correct, however, the result of image2 is the same as image1. May I ask how to solve this problem?
I test the model on Raspberry 3. ncsdk version:2.08
