Transform inference results to images in "Inference SRCNN" notebook

LoSealL / VideoSuperResolution

A collection of state-of-the-art video or single-image super-resolution architectures, reimplemented in tensorflow.

MIT License

1.62k stars 296 forks source link

Transform inference results to images in "Inference SRCNN" notebook #121

Open bmachin opened 3 years ago

bmachin commented 3 years ago

Hi, I'm trying to test some models using the "Inference SRCNN" notebook, and I can't seem to be able to convert the inference results to images.

I've tried using array_to_img from VSR.Util.ImageProcess, but it returns a completely black image. Also tried other approaches, re-scaling to 0-255 and moving the axes around but I'm not getting it right.

How should I process ret[0] to get an actual image?

Thank you!