search

phillipi / pix2pix

Image-to-image translation with conditional adversarial nets
https://phillipi.github.io/pix2pix/
Other
10.11k stars 1.71k forks source link

i want to generate a colorized shoe image using generator. what is the input node name to generator and output node name and their dimensions? #194

Open abhi96624 opened 4 years ago

junyanz commented 4 years ago

For Lua code, please refer to model testing in our Getting Started. You need to put your images in a directory. This is the line for model inference.

abhi96624 commented 4 years ago

thnx @junyanz for such a fast reply :) , actually i was referring to the tensorflow port of this model by https://github.com/affinelayer/pix2pix-tensorflow, i converted the above to pb file using output node as (deprocess_2/truediv - i am not sure if its correct output node) , when i inspected the model using tensorboard the generator(encoder and decoder layer looks good) but generator input node is a 'batch'(which itself contain 2 nodes batch/fifoqueueand batch/n) none of them is a placeholder and fifoqueue dimension is a list. so i dont know how to infer (that is feed an input image into generator and get output image) I am struck now , may be you can help :)

Thanks again for your help :)

I used the pretrained model - Edge2shoe https://github.com/affinelayer/pix2pix-tensorflow

junyanz commented 4 years ago

I see. Sorry that I am not familiar with the TensorFlow repo. Maybe you can post it on that repo.