[Example] image super-resolution

[huningxin]

@GreyZzzzzzXh , SRGAN might be interested by you.

[GreyZzzzzzXh]

@huningxin ,yes SRGAN is in my consideration and i'll try it in the future. This project is a good reference, thanks!

[GreyZzzzzzXh]

1. I trained the SRGAN model on DIV2K dataset using tensorlayer. Reference: tensorlayer/srgan

2. The generator of SRGAN is called SRResNet, being like this. We can achieve a 4x upscaling. According to the origin paper, set B=16.

3. SRResNet uses operation PixelShuffler(DepthToSpace) which is not supported in tflite yet, see TfLite buildin ops. So i use transpose + batchToSpace + transpose instead. New ops we need: transpose, batchToSpace, tanh

4. A simple image demo with WebGL backend. input size [96, 96, 3] => output size [384, 384, 3] It takes several seconds to infer now. Visit https://greyzzzzzzxh.github.io/webml-examples/index.html to test.

[huningxin]

Very nice! We need to investigate the platform API support for the three new ops. BTW, the link doesn't work for me. Would you mind to double check it?

[GreyZzzzzzXh]

the link doesn't work

updated it.

[huningxin]

The link works, however it reports WebGL backend failure to me. Did I miss any thing?

[GreyZzzzzzXh]

Is there any error or warning? I can get results on linux and android via this link.

[huningxin]

The failure disappears now. It's super (for super resolution). 👍

[huningxin]

OpenVINO pre-trained models also contains a SR model (https://software.intel.com/en-us/openvino-toolkit/documentation/pretrained-models). Please take a look.

[GreyZzzzzzXh]

I trained FSRCNN, a small network for super resolution.

New ops we need: sigmoid, maximum, tanh, transpose_conv

Simple image demo with WebGL backend:

Reference: Accelerating the Super-Resolution Convolutional Neural Network

Visit https://greyzzzzzzxh.github.io/webml-examples/index.html to test.

[ibelem]

Move to Feb 2019 milestone for this issue.