search

mjdietzx / SimGAN

Implementation of Apple's Learning from Simulated and Unsupervised Images through Adversarial Training
MIT License
411 stars 101 forks source link

combined_model should train with y_refined rather than y_real #16

Closed caldweln closed 6 years ago

caldweln commented 6 years ago

https://github.com/wayaai/SimGAN/blob/f4920bd97660ae27d90eb74605d6ec3c4c7e4e55/sim-gan.py#L274

Hi,

Nice implementation, after a quick read it looks like the combined model (specifically the discriminator) is trained with y_real as ground truth, although being fed the generator/refiner output... so wouldn't the ground truth be y_refined instead??

I might be mistaken.. can you take a look?

Thanks

caldweln commented 6 years ago

Never mind... found out it's by design. Reason described here: https://github.com/mchablani/deep-learning/blob/master/gan_mnist/Intro_to_GANs_Exercises.ipynb Paragraph: Discriminator and Generator Losses