Need input image support in PyTorch format

[Mohit0928]

The image capability takes an image input in the following format: [batch_size, height, width, channels]. This is the standard format followed by Tensorflow models. There is another format followed by PyTorch models: [batch_size, channels, height, width]. Right now, we only support the Tensorflow format in the rune. We have to change it to support the PyTorch format. There is also one Tensorflow model (craft-text-detector) which support PyTorch format.