Skip to content

Try alternative approach for 2D patch extraction #9

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 3 commits into
base: main
Choose a base branch
from
Open

Try alternative approach for 2D patch extraction #9

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
9d7ba1b
Try alternative approach
jcohenadad Mar 17, 2023
5107b41
Try with EnsureChannelFirst()
jcohenadad Mar 17, 2023
4042dc7
Fixed https://github.com/ivadomed/model_seg_mouse-sc_wm-gm_t1/issues/…
jcohenadad Mar 17, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
54 changes: 54 additions & 0 deletions train.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,6 +22,7 @@
AsDiscreted,
EnsureTyped,
EnsureChannelFirstd,
EnsureChannelFirst,
Compose,
CropForegroundd,
LoadImaged,
Expand Down Expand Up @@ -137,6 +138,59 @@ def wb_mask(bg_img, mask):
for image_name, label_name in zip(train_images_match, train_labels_match)]
# TODO: add check if data empty

# << Trying suggestion: https://github.com/Project-MONAI/MONAI/discussions/5948#discussioncomment-5330042
# Split train/val
train_files, val_files = data_dicts[:-1], data_dicts[-1:]
# Set deterministic training for reproducibility
set_determinism(seed=0)
# # Volume-level transforms for both image and segmentation
# train_transforms = Compose(
# [
# LoadImaged(keys=["image", "label"]),
# EnsureChannelFirstd(keys=["image", "label"]),
# ScaleIntensityd(keys="image"),
# # RandRotate90d(keys=["image", "label"], prob=0.5, spatial_axes=[0, 1]),
# EnsureTyped(keys=["image", "label"]),
# ]
# )
# # 3D dataset with preprocessing transforms
# volume_ds = CacheDataset(data=train_files, transform=train_transforms)
n_samples = 5
sampler = Compose([
LoadImaged(keys=["image", "label"], image_only=True),
EnsureChannelFirstd(keys=["image", "label"]),
RandCropByPosNegLabeld(
keys=["image", "label"],
image_key="image",
label_key="label",
spatial_size=(200, 200, 1),
pos=1,
neg=0,
num_samples=n_samples,
),
])

ds = PatchDataset(data=train_files,
patch_func=sampler,
samples_per_image=n_samples,
transform=None)
check_loader = DataLoader(ds, batch_size=1)
check_data = first(check_loader)

print("First volume's shape: ", check_data["image"].shape, check_data["label"].shape)
i=0
for check_data in check_loader:
if 'image' in check_data:
print(f"{i}: {check_data['image'].size()}")
i += 1

# IGNORE CODE BELOW-- DEBUGGING






# Iterate across image/label 3D volume, fetch non-empty slice and output a single list of image/label pair
patch_data = []
for data_dict in data_dicts:
Expand Down