tiangexiang / bix-nas Goto Github PK

Dear author：

Your team has made great contributions to the development of medical image segmentation, such as BiO-Net, BiX-NAS and so on. 
I have a question about the BiO-Net++. BiO-Net++ adds a lot of dense connections between encode blocks and decode blocks on the basis of BiO-Net. Why is the number of parameters two orders of magnitude less than that of BiO-Net?

Hope to get your answer, thank you very much.

Is the number of out_channel in each block the same

Thanks for your extraordinary work !

At Phase2, it goes well when 15 subnetworks are sampled at iteration 5.
But when it goes to interation 4, the subnetworks tripled as 3 networks in pareto front are obtained in previous iter.
45 subnetworks causes an OOM error in a single NVIDIA GeForce RTX 2080 Ti (11GB) in my case.
How much GPU memory does it run the whole experiment pipeline ?

I am able to reproduce the MoNuSeg mIoU and DICE for phase1 search. Parameters are the same, but the reported MACs in the training log seems 15-20x as the paper shows. It's because MACs are caculated with a smaller input size (resolution) to align with other works ?

best,

Below Replace with the IAA augmented method
transform = torchvision.transforms.Compose([
transforms.Resize((224, 224)),
transforms.ColorJitter(hue=.05, saturation=.05),
transforms.RandomHorizontalFlip(),
transforms.RandomRotation(20, resample=PIL.Image.BILINEAR),
transforms.ToTensor()
])

x, y = self.transforms(images=self.x[None, index], segmentation_maps=self.y[None, index])
TypeError: call() got an unexpected keyword argument 'images'