update 3d_classification (#804)

Signed-off-by: KumoLiu <yunl@nvidia.com>

Author: KumoLiu

3d_classification/densenet_training_array.ipynb

@@ -107,7 +107,6 @@
     "    RandRotate90,\n",
     "    Resize,\n",
     "    ScaleIntensity,\n",
-    "    EnsureType\n",
     ")\n",
     "\n",
     "pin_memory = torch.cuda.is_available()\n",
@@ -210,24 +209,24 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 6,
+   "execution_count": 5,
    "metadata": {},
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "<class 'torch.Tensor'> torch.Size([3, 1, 96, 96, 96]) tensor([[1., 0.],\n",
+      "<class 'monai.data.meta_tensor.MetaTensor'> (3, 1, 96, 96, 96) tensor([[1., 0.],\n",
       "        [1., 0.],\n",
       "        [1., 0.]]) torch.Size([3, 2])\n"
      ]
     }
    ],
    "source": [
     "# Define transforms\n",
-    "train_transforms = Compose([ScaleIntensity(), AddChannel(), Resize((96, 96, 96)), RandRotate90(), EnsureType()])\n",
+    "train_transforms = Compose([ScaleIntensity(), AddChannel(), Resize((96, 96, 96)), RandRotate90()])\n",
     "\n",
-    "val_transforms = Compose([ScaleIntensity(), AddChannel(), Resize((96, 96, 96)), EnsureType()])\n",
+    "val_transforms = Compose([ScaleIntensity(), AddChannel(), Resize((96, 96, 96))])\n",
     "\n",
     "# Define nifti dataset, data loader\n",
     "check_ds = ImageDataset(image_files=images, labels=labels, transform=train_transforms)\n",
@@ -500,9 +499,9 @@
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "Python [conda env:monai]",
+   "display_name": "Python 3 (ipykernel)",
    "language": "python",
-   "name": "conda-env-monai-py"
+   "name": "python3"
   },
   "language_info": {
    "codemirror_mode": {
@@ -514,7 +513,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.9.7"
+   "version": "3.8.13"
   }
  },
  "nbformat": 4,

3d_classification/ignite/densenet_evaluation_array.py

@@ -17,12 +17,11 @@
 import torch
 from ignite.engine import _prepare_batch, create_supervised_evaluator
 from ignite.metrics import Accuracy
-from torch.utils.data import DataLoader
 
 import monai
-from monai.data import ImageDataset
+from monai.data import ImageDataset, DataLoader
 from monai.handlers import CheckpointLoader, ClassificationSaver, StatsHandler
-from monai.transforms import AddChannel, Compose, Resize, ScaleIntensity, EnsureType
+from monai.transforms import EnsureChannelFirst, Compose, Resize, ScaleIntensity
 
 
 def main():
@@ -50,9 +49,9 @@ def main():
     labels = np.array([0, 0, 1, 0, 1, 0, 1, 0, 1, 0], dtype=np.int64)
 
     # define transforms for image
-    val_transforms = Compose([ScaleIntensity(), AddChannel(), Resize((96, 96, 96)), EnsureType()])
+    val_transforms = Compose([ScaleIntensity(), EnsureChannelFirst(), Resize((96, 96, 96))])
     # define image dataset
-    val_ds = ImageDataset(image_files=images, labels=labels, transform=val_transforms, image_only=False)
+    val_ds = ImageDataset(image_files=images, labels=labels, transform=val_transforms, image_only=True)
     # create DenseNet121
     device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
     net = monai.networks.nets.DenseNet121(spatial_dims=3, in_channels=1, out_channels=2).to(device)
@@ -78,7 +77,7 @@ def prepare_batch(batch, device=None, non_blocking=False):
     # for the array data format, assume the 3rd item of batch data is the meta_data
     prediction_saver = ClassificationSaver(
         output_dir="tempdir",
-        batch_transform=lambda batch: batch[2],
+        batch_transform=lambda batch: batch[0].meta,
         output_transform=lambda output: output[0].argmax(1),
     )
     prediction_saver.attach(evaluator)

3d_classification/ignite/densenet_evaluation_dict.py

@@ -17,11 +17,11 @@
 import torch
 from ignite.engine import _prepare_batch, create_supervised_evaluator
 from ignite.metrics import Accuracy
-from torch.utils.data import DataLoader
 
 import monai
+from monai.data import DataLoader
 from monai.handlers import CheckpointLoader, ClassificationSaver, StatsHandler
-from monai.transforms import AddChanneld, Compose, LoadImaged, Resized, ScaleIntensityd, EnsureTyped
+from monai.transforms import Compose, LoadImaged, Resized, ScaleIntensityd
 
 
 def main():
@@ -52,11 +52,9 @@ def main():
     # define transforms for image
     val_transforms = Compose(
         [
-            LoadImaged(keys=["img"]),
-            AddChanneld(keys=["img"]),
+            LoadImaged(keys=["img"], ensure_channel_first=True),
             ScaleIntensityd(keys=["img"]),
             Resized(keys=["img"], spatial_size=(96, 96, 96)),
-            EnsureTyped(keys=["img"]),
         ]
     )
 
@@ -85,7 +83,7 @@ def prepare_batch(batch, device=None, non_blocking=False):
     prediction_saver = ClassificationSaver(
         output_dir="tempdir",
         name="evaluator",
-        batch_transform=lambda batch: batch["img_meta_dict"],
+        batch_transform=lambda batch: batch["img"].meta,
         output_transform=lambda output: output[0].argmax(1),
     )
     prediction_saver.attach(evaluator)

3d_classification/ignite/densenet_training_array.py

@@ -18,12 +18,11 @@
 from ignite.engine import Events, create_supervised_evaluator, create_supervised_trainer
 from ignite.handlers import EarlyStopping, ModelCheckpoint
 from ignite.metrics import Accuracy
-from torch.utils.data import DataLoader
 
 import monai
-from monai.data import ImageDataset, decollate_batch
+from monai.data import ImageDataset, decollate_batch, DataLoader
 from monai.handlers import StatsHandler, TensorBoardStatsHandler, stopping_fn_from_metric
-from monai.transforms import AddChannel, Compose, RandRotate90, Resize, ScaleIntensity, EnsureType
+from monai.transforms import EnsureChannelFirst, Compose, RandRotate90, Resize, ScaleIntensity
 
 
 def main():
@@ -61,8 +60,8 @@ def main():
     labels = np.array([0, 0, 0, 1, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 0], dtype=np.int64)
 
     # define transforms
-    train_transforms = Compose([ScaleIntensity(), AddChannel(), Resize((96, 96, 96)), RandRotate90(), EnsureType()])
-    val_transforms = Compose([ScaleIntensity(), AddChannel(), Resize((96, 96, 96)), EnsureType()])
+    train_transforms = Compose([ScaleIntensity(), EnsureChannelFirst(), Resize((96, 96, 96)), RandRotate90()])
+    val_transforms = Compose([ScaleIntensity(), EnsureChannelFirst(), Resize((96, 96, 96))])
 
     # define image dataset, data loader
     check_ds = ImageDataset(image_files=images, labels=labels, transform=train_transforms)

3d_classification/ignite/densenet_training_dict.py

@@ -17,12 +17,11 @@
 import torch
 from ignite.engine import Events, _prepare_batch, create_supervised_evaluator, create_supervised_trainer
 from ignite.handlers import EarlyStopping, ModelCheckpoint
-from torch.utils.data import DataLoader
 
 import monai
-from monai.data import decollate_batch
+from monai.data import decollate_batch, DataLoader
 from monai.handlers import ROCAUC, StatsHandler, TensorBoardStatsHandler, stopping_fn_from_metric
-from monai.transforms import Activations, AddChanneld, AsDiscrete, Compose, LoadImaged, RandRotate90d, Resized, ScaleIntensityd, EnsureTyped, EnsureType
+from monai.transforms import Activations, AsDiscrete, Compose, LoadImaged, RandRotate90d, Resized, ScaleIntensityd
 
 
 def main():
@@ -64,21 +63,17 @@ def main():
     # define transforms for image
     train_transforms = Compose(
         [
-            LoadImaged(keys=["img"]),
-            AddChanneld(keys=["img"]),
+            LoadImaged(keys=["img"], ensure_channel_first=True),
             ScaleIntensityd(keys=["img"]),
             Resized(keys=["img"], spatial_size=(96, 96, 96)),
             RandRotate90d(keys=["img"], prob=0.8, spatial_axes=[0, 2]),
-            EnsureTyped(keys=["img"]),
         ]
     )
     val_transforms = Compose(
         [
-            LoadImaged(keys=["img"]),
-            AddChanneld(keys=["img"]),
+            LoadImaged(keys=["img"], ensure_channel_first=True),
             ScaleIntensityd(keys=["img"]),
             Resized(keys=["img"], spatial_size=(96, 96, 96)),
-            EnsureTyped(keys=["img"]),
         ]
     )
 
@@ -126,8 +121,8 @@ def prepare_batch(batch, device=None, non_blocking=False):
     # add evaluation metric to the evaluator engine
     val_metrics = {metric_name: ROCAUC()}
 
-    post_label = Compose([EnsureType(), AsDiscrete(to_onehot=2)])
-    post_pred = Compose([EnsureType(), Activations(softmax=True)])
+    post_label = Compose([AsDiscrete(to_onehot=2)])
+    post_pred = Compose([Activations(softmax=True)])
     # Ignite evaluator expects batch=(img, label) and returns output=(y_pred, y) at every iteration,
     # user can add output_transform to return other values
     evaluator = create_supervised_evaluator(net, val_metrics, device, True, prepare_batch=prepare_batch, output_transform=lambda x, y, y_pred: ([post_pred(i) for i in decollate_batch(y_pred)], [post_label(i) for i in decollate_batch(y)]))

3d_classification/torch/densenet_evaluation_array.py

@@ -15,11 +15,10 @@
 
 import numpy as np
 import torch
-from torch.utils.data import DataLoader
 
 import monai
-from monai.data import CSVSaver, ImageDataset
-from monai.transforms import AddChannel, Compose, Resize, ScaleIntensity, EnsureType
+from monai.data import CSVSaver, ImageDataset, DataLoader
+from monai.transforms import AddChannel, Compose, Resize, ScaleIntensity
 
 
 def main():
@@ -47,10 +46,10 @@ def main():
     labels = np.array([0, 0, 1, 0, 1, 0, 1, 0, 1, 0], dtype=np.int64)
 
     # Define transforms for image
-    val_transforms = Compose([ScaleIntensity(), AddChannel(), Resize((96, 96, 96)), EnsureType()])
+    val_transforms = Compose([ScaleIntensity(), AddChannel(), Resize((96, 96, 96))])
 
     # Define image dataset
-    val_ds = ImageDataset(image_files=images, labels=labels, transform=val_transforms, image_only=False)
+    val_ds = ImageDataset(image_files=images, labels=labels, transform=val_transforms, image_only=True)
     # create a validation data loader
     val_loader = DataLoader(val_ds, batch_size=2, num_workers=4, pin_memory=torch.cuda.is_available())
 
@@ -70,7 +69,7 @@ def main():
             value = torch.eq(val_outputs, val_labels)
             metric_count += len(value)
             num_correct += value.sum().item()
-            saver.save_batch(val_outputs, val_data[2])
+            saver.save_batch(val_outputs, val_images.meta)
         metric = num_correct / metric_count
         print("evaluation metric:", metric)
         saver.finalize()

3d_classification/torch/densenet_evaluation_dict.py

@@ -15,11 +15,10 @@
 
 import numpy as np
 import torch
-from torch.utils.data import DataLoader
 
 import monai
-from monai.data import CSVSaver
-from monai.transforms import AddChanneld, Compose, LoadImaged, Resized, ScaleIntensityd, EnsureTyped
+from monai.data import CSVSaver, DataLoader
+from monai.transforms import AddChanneld, Compose, LoadImaged, Resized, ScaleIntensityd
 
 
 def main():
@@ -54,7 +53,6 @@ def main():
             AddChanneld(keys=["img"]),
             ScaleIntensityd(keys=["img"]),
             Resized(keys=["img"], spatial_size=(96, 96, 96)),
-            EnsureTyped(keys=["img"]),
         ]
     )
 
@@ -78,7 +76,7 @@ def main():
             value = torch.eq(val_outputs, val_labels)
             metric_count += len(value)
             num_correct += value.sum().item()
-            saver.save_batch(val_outputs, val_data["img_meta_dict"])
+            saver.save_batch(val_outputs, val_data["img"].meta)
         metric = num_correct / metric_count
         print("evaluation metric:", metric)
         saver.finalize()

3d_classification/torch/densenet_training_array.py

@@ -15,12 +15,11 @@
 
 import numpy as np
 import torch
-from torch.utils.data import DataLoader
 from torch.utils.tensorboard import SummaryWriter
 
 import monai
-from monai.data import ImageDataset
-from monai.transforms import AddChannel, Compose, RandRotate90, Resize, ScaleIntensity, EnsureType
+from monai.data import ImageDataset, DataLoader
+from monai.transforms import AddChannel, Compose, RandRotate90, Resize, ScaleIntensity
 
 
 def main():
@@ -58,8 +57,8 @@ def main():
     labels = np.array([0, 0, 0, 1, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 0], dtype=np.int64)
 
     # Define transforms
-    train_transforms = Compose([ScaleIntensity(), AddChannel(), Resize((96, 96, 96)), RandRotate90(), EnsureType()])
-    val_transforms = Compose([ScaleIntensity(), AddChannel(), Resize((96, 96, 96)), EnsureType()])
+    train_transforms = Compose([ScaleIntensity(), AddChannel(), Resize((96, 96, 96)), RandRotate90()])
+    val_transforms = Compose([ScaleIntensity(), AddChannel(), Resize((96, 96, 96))])
 
     # Define image dataset, data loader
     check_ds = ImageDataset(image_files=images, labels=labels, transform=train_transforms)

3d_classification/torch/densenet_training_dict.py

@@ -15,13 +15,12 @@
 
 import numpy as np
 import torch
-from torch.utils.data import DataLoader
 from torch.utils.tensorboard import SummaryWriter
 
 import monai
-from monai.data import decollate_batch
+from monai.data import decollate_batch, DataLoader
 from monai.metrics import ROCAUCMetric
-from monai.transforms import Activations, AddChanneld, AsDiscrete, Compose, LoadImaged, RandRotate90d, Resized, ScaleIntensityd, EnsureTyped, EnsureType
+from monai.transforms import Activations, AddChanneld, AsDiscrete, Compose, LoadImaged, RandRotate90d, Resized, ScaleIntensityd
 
 
 def main():
@@ -68,7 +67,6 @@ def main():
             ScaleIntensityd(keys=["img"]),
             Resized(keys=["img"], spatial_size=(96, 96, 96)),
             RandRotate90d(keys=["img"], prob=0.8, spatial_axes=[0, 2]),
-            EnsureTyped(keys=["img"]),
         ]
     )
     val_transforms = Compose(
@@ -77,11 +75,10 @@ def main():
             AddChanneld(keys=["img"]),
             ScaleIntensityd(keys=["img"]),
             Resized(keys=["img"], spatial_size=(96, 96, 96)),
-            EnsureTyped(keys=["img"]),
         ]
     )
-    post_pred = Compose([EnsureType(), Activations(softmax=True)])
-    post_label = Compose([EnsureType(), AsDiscrete(to_onehot=2)])
+    post_pred = Compose([Activations(softmax=True)])
+    post_label = Compose([AsDiscrete(to_onehot=2)])
 
     # Define dataset, data loader
     check_ds = monai.data.Dataset(data=train_files, transform=train_transforms)