osandvold302 / trabecular_bone_unet Goto Github PK

undersampled_sample = np.multiply(np.reshape(full_kspace[counter, 0, :, :, :], (num_slices, img_height, img_width)), mask_2d)
undersampled_kspace[counter, 0, :, :, :] = np.reshape(undersampled_sample, (img_height, img_width, num_slices))

reference code location

@brandonclintonjones when undersampling the 3d volume kspace, does this reshaping seem appropriate?

to recall: full_kspace would be (512,512,60) and mask_2d.shape = (512,512). The undersampled_kspace has resulting size (num_scans, 1, 512, 512, 60)

To do some quick prototyping/testing I've been creating the conda environment with the environment.yml file on the add-3d-option branch and setting max_iter =1 for generating the pdf.

@brandonclintonjones I'm not going to assign you to this issue, because then I think you'll get updates every time a push occurs? Or if I'm just using this space to think out loud. So I'll try to ping you if there's something specific I'm asking.

Here's a few questions right off the bat on 3D implementation:

Assumptions about generated undersampled kspace

• Because we'll be needing 2 different kspace undersampled "images" for 3D, what can we assume about the optimization for finding the least interference for the resulting volume?

• Does the pdf need to change in both directions or just the depth direction?

Fails to build cnn on lsni2

2020-08-02 16:23:13.432065: W tensorflow/core/framework/op_kernel.cc:1273] OP_REQUIRES failed at conv_ops_3d.cc:161 : Resource exhausted: OOM when allocating tensor with shape[25,32,512,512,64] and type float on /job:localhost/replica:0/task:0/device:GPU:0 by allocator GPU_0_bfc

But I'm pretty sure it's using the new GPU:

Found device 0 with properties:
name: Tesla V100-PCIE-32GB major: 7 minor: 0 memoryClockRate(GHz): 1.38
pciBusID: 0000:d8:00.0
2020-08-02 16:27:00.998709: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1115] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 30511 MB memory) -> physical GPU (device: 0, name: Tesla V100-PCIE-32GB, pci bus id: 0000:d8:00.0, compute capability: 7.0)

I decreased my batch size to 3. Only using the 9 layers (from this post it seems like that's well below concern).

Attached full output