Does SNPE support variable x,y dimensions as input to a conv2d layer?
For example, in keras you can specify input_shape=(None, None, 1) as a Conv2D parameter.
If it is supported, how would the input shape be defined when using snpe-tensorflow-to-dlc? Can I specify something like,
snpe-tensorflow-to-dlc ... --input_dim conv2d_1_input 1,None,None,1 ...