I used your publication for ICLR 2019 Reproducibility Challenge and here's my tensorflow implementation:
https://github.com/KleinYuan/tf-stacked-unets
It works well on CIFAR-10/100 however for ImageNet task-3, it does not seem converging.
Could you share more details on that part? like the training curve?
(ImageNet task1-2 are too large and I am afraid that I could not afford it. Therefore, I could not really run your training code on imagenet)
I used your publication for ICLR 2019 Reproducibility Challenge and here's my tensorflow implementation:
https://github.com/KleinYuan/tf-stacked-unets
It works well on CIFAR-10/100 however for ImageNet task-3, it does not seem converging.
Could you share more details on that part? like the training curve?
(ImageNet task1-2 are too large and I am afraid that I could not afford it. Therefore, I could not really run your training code on imagenet)