leohsuofnthu / pytorch-iterativefcn Goto Github PK

Hello, Thanks for your great work. The code is very neat and readable, from which I learned a lot.
But I have some questions while implementing this repo on csi2014 dataset.

1. I set batch_size to 64(model's num_channels=64) but only consumes about 4500Mbit in my GPU(GTX 1080). You said you set batch_size to 1 due to memory limitation. Do you think there is something wrong about my setting considering my low memory use?
2. In utils.py->force_inside_img(func), you set coordinate inside the image using img_shape[2]. I think it is more reasonable to limit z by img_shape[0], y by img_shape[1] and x by img_shape[2]. I'm not sure if it should be like that.
   Very grateful if you can give me some guide according to my question.

Could you please provide us with the dataset?
My email is [email protected].
Thanks

Hello Leohsuofnthu
I am very interested by your work and would like to run your code.
However, the dataset of the spine segmentation challenge in CSI2014 is no more available : http://spineweb.digitalimaginggroup.ca/spineweb/index.php?n=Main.Datasets#Dataset_2.3A_Spine_and_Vertebrae_Segmentation

It seems that the server is no more available.

Would you be kind enough to upload the dataset on Github or send it to me ([email protected]) ?
Kind regards,
JM

How can we get 2d visual result from 3d .nrrd or 3d .mhd?

In train.py, you use next(iter(train_loader)) to fetch the patches every iteration.

However , in such way you build an iterator of train_loader every iteration,and next() would return the first element of the iterator every time, which means during the whole iteration there is only one subject is used to training.

I'm just confused about this.Could anyone give me some explanation or discussion?

Why use sigmoid instead of softmax in segmentation network?

i think this is multi-class segmentation problem......

Hi CHIH CHAO,
When I'm testing the code (instance_segment.py), I found the program will step into an endless loop when segment images (case6, case9...) .
There are two situations in converge iterations:
1.The Modified centre is always the same as prev centre and will not break out of the loop
2.when iters=lim_alternate_times, converge ,update ins and mask, to this point, z_up may exceed the image shape, cause the program to crash. Why does this happen after doing correction to be in-frame?
Looking forward to hearing from you, thank you~

Why？