python，神经网络，要使用的input_shape的尺寸和值

Question

i need to build a CNN model on a dataset which has 65536 rows (represent 1 image for each), 49 columns (7x7 image) and binary class(50th column).

I am referencing examples on performing CNN using mnist dataset but i failed to build the train model.

When i'm on this line of code:

model.add(Conv2D(30 ,(5,5), padding='valid',  activation='relu',input_shape=(1,7,7))

i am have this error :

ValueError: Negative dimension size caused by subtracting 5 from 1 for 'conv2d_42/convolution' (op: 'Conv2D') with input shapes: [?,1,7,7], [5,5,7,30].

where i try this :

model.add(Conv2D(30 ,(5,5), padding='valid',  activation='relu',input_shape=(1(7,7)))

I had this :

TypeError: int() argument must be a string or a number, not 'tuple'

what im asking is what value of input_shape i should be using to build the model

Answer 1

As you know, Keras can run on top of either Theano of Tensorflow. You are using Theano dimension ordering (channels, height, width) but your Keras seems to be using Tensorflow backend, with Tensorflow dimension ordering (height, width, channels).

I would suggest rewriting the code and put channel dimension (=1) last. There are also way of changing backend and/or dimension ordering by editing the keras.json.