initial dints tutorials (Project-MONAI#449)

* initialize dints tutorials Signed-off-by: dongy <dongy@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update scripts Signed-off-by: dongy <dongy@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update scripts Signed-off-by: dongy <dongy@nvidia.com> * update scripts Signed-off-by: dongy <dongy@nvidia.com> * Update README.md * Update README.md * update scripts Signed-off-by: dongy <dongy@nvidia.com> * Change Dints interface * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Modify scripts for new Dints interface * update scripts Signed-off-by: dongy <dongy@nvidia.com> * update scripts Signed-off-by: dongy <dongy@nvidia.com> * update scripts Signed-off-by: dongy <dongy@nvidia.com> * update scripts Signed-off-by: dongy <dongy@nvidia.com> * Test push * update scripts Signed-off-by: dongy <dongy@nvidia.com> * update scripts Signed-off-by: dongy <dongy@nvidia.com> * update scripts Signed-off-by: dongy <dongy@nvidia.com> * update readme Signed-off-by: Dong Yang <dongy@dongy-mlt.client.nvidia.com> * Add readme * Update readme * Change lr in search * update readme Signed-off-by: Dong Yang <dongy@dongy-mlt.client.nvidia.com> * Enable single GPU * update readme Signed-off-by: Dong Yang <dongy@dongy-mlt.client.nvidia.com> * Add visualization tutorial transform image (Project-MONAI#448) * [DLMED] add visualization tutorial Signed-off-by: Nic Ma <nma@nvidia.com> * [DLMED] add image Signed-off-by: Nic Ma <nma@nvidia.com> * [DLMED] update notebook Signed-off-by: Nic Ma <nma@nvidia.com> * [DLMED] update notebook Signed-off-by: Nic Ma <nma@nvidia.com> * [DLMED] add reference image Signed-off-by: Nic Ma <nma@nvidia.com> * [DLMED] fix PEP error Signed-off-by: Nic Ma <nma@nvidia.com> * Update spleen_segmentation_3d.ipynb (Project-MONAI#455) The `pip install` statement is missing pytorch-ignite. Changing: `!python -c "import monai" || pip install -q "monai-weekly[gdown, nibabel, tqdm]"` to `!python -c "import monai" || pip install -q "monai-weekly[gdown, nibabel, tqdm, ignite]"` * Figures added, pretrained weights link added, minor fixes (Project-MONAI#456) * Figures added, pretrained weights link added, minor fixes Signed-off-by: vnath <vnath@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci Co-authored-by: vnath <vnath@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Add itkwidgets example in notebook (Project-MONAI#454) * [DLMED] add itkwidgets Signed-off-by: Nic Ma <nma@nvidia.com> * [DLMED] add reference screenshot Signed-off-by: Nic Ma <nma@nvidia.com> * [DLMED] update to PLS orientation Signed-off-by: Nic Ma <nma@nvidia.com> * MIL example (Project-MONAI#431) * MIL example Signed-off-by: myron <amyronenko@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * mil tutorial update Signed-off-by: myron <amyronenko@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * mil tutorial update Signed-off-by: myron <amyronenko@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * small updates Signed-off-by: myron <amyronenko@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * updated images Signed-off-by: myron <amyronenko@nvidia.com> * gdown for json Signed-off-by: myron <amyronenko@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Update README Signed-off-by: Behrooz <3968947+drbeh@users.noreply.github.com> * Fix formatting and typos Signed-off-by: Behrooz <3968947+drbeh@users.noreply.github.com> * small fixes Signed-off-by: myron <amyronenko@nvidia.com> * stats Signed-off-by: myron <amyronenko@nvidia.com> * pip install Signed-off-by: myron <amyronenko@nvidia.com> * README fixes Signed-off-by: myron <amyronenko@nvidia.com> Co-authored-by: am <am> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Behrooz <3968947+drbeh@users.noreply.github.com> Co-authored-by: Nic Ma <nma@nvidia.com> * 450 update AsDiscrete (Project-MONAI#451) * update asdiscrete Signed-off-by: Yiheng Wang <vennw@nvidia.com> * update postprocessing figures Signed-off-by: Yiheng Wang <vennw@nvidia.com> * fix version error of mutual info Signed-off-by: Yiheng Wang <vennw@nvidia.com> * update to use include in torchin Signed-off-by: Yiheng Wang <vennw@nvidia.com> * 459 update nvidia flare 2.0 example (Project-MONAI#460) * update nvidia flare 2.0 example Signed-off-by: Yiheng Wang <vennw@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * enhance code format Signed-off-by: Yiheng Wang <vennw@nvidia.com> * update folder and readme Signed-off-by: Yiheng Wang <vennw@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Weights Link Updated (Project-MONAI#465) Signed-off-by: vnath <vnath@nvidia.com> Co-authored-by: vnath <vnath@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fixes readme typos Signed-off-by: Wenqi Li <wenqil@nvidia.com> * update readmes Signed-off-by: Wenqi Li <wenqil@nvidia.com> * update readme Signed-off-by: Wenqi Li <wenqil@nvidia.com> * qa commit Signed-off-by: Wenqi Li <wenqil@nvidia.com> * link Signed-off-by: Wenqi Li <wenqil@nvidia.com> Co-authored-by: dongy <dongy@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: heyufan1995 <heyufan1995@gmail.com> Co-authored-by: Dong Yang <dongy@dongy-mlt.client.nvidia.com> Co-authored-by: Nic Ma <nma@nvidia.com> Co-authored-by: nvahmadi <85887144+nvahmadi@users.noreply.github.com> Co-authored-by: Vishwesh <vishwesh.nath@gmail.com> Co-authored-by: vnath <vnath@nvidia.com> Co-authored-by: myron <myron@users.noreply.github.com> Co-authored-by: Behrooz <3968947+drbeh@users.noreply.github.com> Co-authored-by: Yiheng Wang <68361391+yiheng-wang-nv@users.noreply.github.com> Co-authored-by: Wenqi Li <wenqil@nvidia.com>
boneseva · Dec 1, 2021 · 9b69852 · 9b69852
1 parent 38d1e47
commit 9b69852
Show file tree

Hide file tree

Showing 13 changed files with 1,514 additions and 3 deletions.
diff --git a/automl/DiNTS/Figures/arch_ram-cost-0.8.png b/automl/DiNTS/Figures/arch_ram-cost-0.8.png
diff --git a/automl/DiNTS/Figures/search_space.png b/automl/DiNTS/Figures/search_space.png
diff --git a/automl/DiNTS/Figures/training_loss.png b/automl/DiNTS/Figures/training_loss.png
diff --git a/automl/DiNTS/Figures/validation_metric.png b/automl/DiNTS/Figures/validation_metric.png
diff --git a/automl/DiNTS/README.md b/automl/DiNTS/README.md
@@ -0,0 +1,76 @@
+# Examples of DiNTS: Differentiable neural network topology search
+
+In this tutorial, we present a novel neural architecture search algorithm for 3D medical image segmentation. The datasets used in this tutorial are Task07 Pancreas (CT images) and Task09 Spleen (CT images) from [Medical Segmentation Decathlon](http://medicaldecathlon.com/). The implementation is based on:
+
+Yufan He, Dong Yang, Holger Roth, Can Zhao, Daguang Xu: "[DiNTS: Differentiable Neural Network Topology Search for 3D Medical Image Segmentation.](https://openaccess.thecvf.com/content/CVPR2021/papers/He_DiNTS_Differentiable_Neural_Network_Topology_Search_for_3D_Medical_Image_CVPR_2021_paper.pdf)" In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 5841-5850. 2021.
+
+![0.8](./Figures/arch_ram-cost-0.8.png)
+![space](./Figures/search_space.png)
+
+## Requirements
+The script is tested with:
+- `Ubuntu 20.04` and  `CUDA 11`
+- The searching and training stage requires at least two 16GB GPUs.
+
+## Dependencies and installation
+### Download and install Nvidia PyTorch Docker
+```bash
+docker pull nvcr.io/nvidia/pytorch:21.10-py3
+```
+### Download the repository
+```bash
+git clone https://github.com/Project-MONAI/tutorials.git
+```
+### Run into Docker
+```
+sudo docker run -it --gpus all --pid=host --shm-size 16G -v /location/to/tutorials/automl/DiNTS/:/workspace/DiNTS/  nvcr.io/nvidia/pytorch:21.10-py3
+```
+### Install MONAI and dependencies
+```bash
+bash install.sh
+```
+
+## Data
+[Spleen CT dataset](https://drive.google.com/drive/folders/1HqEgzS8BV2c7xYNrZdEAnrHk7osJJ--2) and [Pancreas MRI dataset](https://drive.google.com/drive/folders/1HqEgzS8BV2c7xYNrZdEAnrHk7osJJ--2)
+from [Medical Segmentation Decathlon](http://medicaldecathlon.com/) is used. You can manually download it and save it to args.root. Otherwise, the script will automatic
+download the dataset.
+
+## Examples
+The tutorial contains two stages: searching stage and training stage. An architecture is searched and saved into a `.pth` file using `search_dints.py`.
+The searched architecture will be loaded by `train_dints.py` and re-trained for spleen segmentation.
+
+Check all possible options:
+```bash
+cd ./DiNTS/
+python search_dints.py -h
+python train_dints.py -h
+```
+
+### Searching
+- Add the following script to the commands of running into docker (optional)
+```
+-v /your_downloaded_data_root/Task07_Pancreas/:/workspace/data_msd/Task07_Pancreas/
+```
+- Change ``NUM_GPUS_PER_NODE`` to your number of GPUs.
+- Run `bash search_dints.sh`
+
+### Training
+- Add the following script to the commands of running into docker (Optional)
+```
+-v /your_downloaded_data_root/Task09_Spleen/:/workspace/data_msd/Task09_Spleen/
+```
+- Change ``ARCH_CKPT`` to point to the architecture file (.pth) from the searching stage.
+- Change ``NUM_GPUS_PER_NODE`` to your number of GPUs.
+- Run `bash train_dints.sh`
+
+Training loss and validation metric curves are shown as follows. The experiments utilized 8 NVIDIA A100 GPUs.
+
+![training_loss](./Figures/training_loss.png)
+
+![validation_metric](./Figures/validation_metric.png)
+
+## Questions and bugs
+
+- For questions relating to the use of MONAI, please use our [Discussions tab](https://github.com/Project-MONAI/MONAI/discussions) on the main repository of MONAI.
+- For bugs relating to MONAI functionality, please create an issue on the [main repository](https://github.com/Project-MONAI/MONAI/issues).
+- For bugs relating to the running of a tutorial, please create an issue in [this repository](https://github.com/Project-MONAI/Tutorials/issues).
diff --git a/automl/DiNTS/arch_code.pth b/automl/DiNTS/arch_code.pth
diff --git a/automl/DiNTS/install.sh b/automl/DiNTS/install.sh
@@ -0,0 +1,12 @@
+#!/bin/bash
+clear
+
+pip install nibabel
+pip install pandas
+
+# Update pip
+python -m pip install -U pip
+# Install scikit-image
+python -m pip install -U scikit-image
+
+pip install git+https://github.com/Project-MONAI/MONAI#egg=monai