search

NVIDIA / pix2pixHD

Synthesizing and manipulating 2048x1024 images with conditional GANs
https://tcwang0509.github.io/pix2pixHD/
Other
6.57k stars 1.38k forks source link

Image to label using pix2pixHD #160

Open jpnavarro-nv opened 4 years ago

jpnavarro-nv commented 4 years ago

What do I need to do in order to use the pix2pixHD code to generate a semantic segmentation map as output (discrete classes, or labels)? In this way, it would be an image-to-label problem. Thanks!

cszer commented 4 years ago

You need: 1.prepare real and segmentation label images 2.distinguish what colors represent particular class 3.train the model 4.use function that can change pixel values which have different color than particular label to 0 5.generate n binary mask using this function this all,try it

yan123yan commented 3 years ago

You need: 1.prepare real and segmentation label images 2.distinguish what colors represent particular class 3.train the model 4.use function that can change pixel values which have different color than particular label to 0 5.generate n binary mask using this function this all,try it

Hi, your answer is very helpful. What do I need to do to solve a label-to-image problem with pix2pixHD? Thanks. (for example, number 3 to a image contains in MNIST)