fully-run notebooks links and fixed contrib voxel models

README.md

@@ -21,8 +21,8 @@ If you run into any problems, chances are your problem has already been solved i
 ### Pre-run notebooks
 
 Notebooks stored in the repository have output intentionally displaced - you can find full auto-generated versions of the notebooks here:
-- [HRNet Penobscot demo](https://deepseismicstore.blob.core.windows.net/shared/HRNet_Penobscot_demo_notebook.ipynb?sp=r&st=2019-12-13T05:11:15Z&se=2019-12-13T13:11:15Z&spr=https&sv=2019-02-02&sr=b&sig=UFwueAhZcJn8o7g1nzD4GGS7lKv9lHIJXJb%2B7kbyUZc%3D)
-- [Dutch F3 dataset](https://deepseismicstore.blob.core.windows.net/shared/F3_block_training_and_evaluation_local.ipynb?sp=r&st=2019-12-13T05:10:17Z&se=2019-12-13T13:10:17Z&spr=https&sv=2019-02-02&sr=b&sig=UUDCulVUddhqfxN0%2FqcdLZ7DmcGnIYk0j%2BlM0EN8WiM%3D)
+- **HRNet Penobscot demo**: [[HTML](https://deepseismicstore.blob.core.windows.net/shared/HRNet_Penobscot_demo_notebook.html)] [[.ipynb](https://deepseismicstore.blob.core.windows.net/shared/HRNet_Penobscot_demo_notebook.ipynb)]
+- **Dutch F3 dataset**: [[HTML](https://deepseismicstore.blob.core.windows.net/shared/F3_block_training_and_evaluation_local.html)] [[.ipynb](https://deepseismicstore.blob.core.windows.net/shared/F3_block_training_and_evaluation_local.ipynb)]
 
 ### Azure Machine Learning
 [Azure Machine Learning](https://docs.microsoft.com/en-us/azure/machine-learning/) enables you to train and deploy your machine learning models and pipelines at scale, ane leverage open-source Python frameworks, such as PyTorch, TensorFlow, and scikit-learn. If you are looking at getting started with using the code in this repository with Azure Machine Learning, refer to [Azure Machine Learning How-to](https://github.com/Azure/MachineLearningNotebooks/tree/master/how-to-use-azureml) to get started.

contrib/experiments/interpretation/dutchf3_voxel/README.md

@@ -1,15 +1,17 @@
-First, make sure that `/mnt/dutch_f3` folder exists and you have write access.
+First, make sure that `${HOME}/data/dutch_f3` folder exists and you have write access.
 
 Next, to get the main input dataset which is the [Dutch F3 dataset](https://terranubis.com/datainfo/Netherlands-Offshore-F3-Block-Complete), 
 navigate to [MalenoV](https://github.com/bolgebrygg/MalenoV) project website and follow the links (which will lead to 
 [this](https://drive.google.com/drive/folders/0B7brcf-eGK8CbGhBdmZoUnhiTWs) download). Save this file as 
-`/mnt/dutch_f3/data.segy`
+`${HOME}/data/dutch_f3/data.segy`
 
 To download the train and validation masks, from the root of the repo, run
 ```bash
-./scripts/get_F3_voxel.sh /mnt/dutch_f3
+./contrib/scripts/get_F3_voxel.sh ${HOME}/data/dutch_f3
 ```
 
 This will also download train and validation masks to the same location as data.segy.
 
 That's it!
+
+To run the training script, run `python train.py --cfg=configs/texture_net.yaml`.

contrib/experiments/interpretation/dutchf3_voxel/configs/texture_net.yaml

@@ -15,7 +15,7 @@ WINDOW_SIZE: 65
 
 DATASET:
   NUM_CLASSES: 2
-  ROOT: /mnt/dutchf3
+  ROOT: /home/maxkaz/data/dutchf3
   FILENAME: data.segy  
 
 MODEL:

contrib/experiments/voxel2pixel/README.md → contrib/experiments/interpretation/voxel2pixel/README.md

@@ -10,15 +10,19 @@ There is also an
 EAGE E-lecture which you can watch: [*Seismic interpretation with deep learning*](https://www.youtube.com/watch?v=lm85Ap4OstM) (YouTube)
 
 ### Setup to get started
-- make sure you follow `SETUP.md` file in root of repo to install all the proper dependencies.
-- download the data by running `scrips/get_F3_voxel.sh` from the root of this repo, i.e. from DeepSeismic folder.
+- make sure you follow `README.md` file in root of repo to install all the proper dependencies.
+- downgrade TensorFlow and pyTorch's CUDA:
+    - downgrade TensorFlow by running `pip install tensorflow-gpu==1.14` 
+    - make sure pyTorch uses downgraded CUDA `pip install torch==1.3.1+cu92 torchvision==0.4.2+cu92 -f https://download.pytorch.org/whl/torch_stable.html`
+- download the data by running `contrib/scrips/get_F3_voxel.sh` from the `contrib` folder of this repo.
 This will download the training and validation labels/masks.
 - to get the main input dataset which is the [Dutch F3 dataset](https://terranubis.com/datainfo/Netherlands-Offshore-F3-Block-Complete), 
 navigate to [MalenoV](https://github.com/bolgebrygg/MalenoV) project website and follow the links (which will lead to 
 [this](https://drive.google.com/drive/folders/0B7brcf-eGK8CbGhBdmZoUnhiTWs) download). Save this file as 
 `interpretation/voxel2pixel/F3/data.segy`
 
-
+If you want to revert downgraded packages, just run `conda env update -f environment/anaconda/local/environment.yml` from the root folder of the repo.
+
 ### Monitoring progress with TensorBoard
 - from the `voxel2pixel` directory, run `tensorboard --logdir='log'` (all runtime logging information is
 written to the `log` folder <br />

contrib/experiments/voxel2pixel/test_parallel.py → contrib/experiments/interpretation/voxel2pixel/test_parallel.py

@@ -7,7 +7,7 @@
 import os
 
 # set default number of GPUs which are discoverable
-N_GPU = 1
+N_GPU = 4
 DEVICE_IDS = list(range(N_GPU))
 os.environ["CUDA_VISIBLE_DEVICES"] = ",".join([str(x) for x in DEVICE_IDS])
 
@@ -245,7 +245,7 @@ def main_worker(gpu, ngpus_per_node, args):
 
 
 parser = argparse.ArgumentParser(description="Seismic Distributed Scoring")
-parser.add_argument("-d", "--data", default="/mnt/dutchf3", type=str, help="default dataset folder name")
+parser.add_argument("-d", "--data", default="/home/maxkaz/data/dutchf3", type=str, help="default dataset folder name")
 parser.add_argument(
     "-s",
     "--slice",

contrib/experiments/voxel2pixel/train.py → contrib/experiments/interpretation/voxel2pixel/train.py

@@ -15,7 +15,7 @@
 import utils
 
 # Parameters
-ROOT_PATH = "/mnt/dutchf3"
+ROOT_PATH = "/home/maxkaz/data/dutchf3"
 INPUT_VOXEL = "data.segy"
 TRAIN_MASK = "inline_339.png"
 VAL_MASK = "inline_405.png"

scripts/get_F3_voxel.sh → contrib/scripts/get_F3_voxel.sh

@@ -2,19 +2,22 @@
 # Copyright (c) Microsoft Corporation. All rights reserved.
 # Licensed under the MIT License.
 
-echo "Downloading Dutch F3 from https://github.com/bolgebrygg/MalenoV"
+echo "Make sure you also download Dutch F3 data from https://github.com/bolgebrygg/MalenoV"
 # fetch Dutch F3 from Malenov project.
 # wget https://drive.google.com/open?id=0B7brcf-eGK8CUUZKLXJURFNYeXM -O interpretation/voxel2pixel/F3/data.segy
 
 if [ $# -eq 0 ]
 then
-    downdirtrain='interpretation/voxel2pixel/F3/train'
-    downdirval='interpretation/voxel2pixel/F3/val'
+    downdirtrain='experiments/interpretation/voxel2pixel/F3/train'
+    downdirval='experiments/interpretation/voxel2pixel/F3/val'
 else
     downdirtrain=$1
     downdirval=$1
 fi
 
+mkdir -p ${downdirtrain}
+mkdir -p ${downdirval}
+
 echo "Downloading train label to $downdirtrain and validation label to $downdirval"
 wget https://github.com/waldeland/CNN-for-ASI/raw/master/F3/train/inline_339.png -O ${downdirtrain}/inline_339.png
 wget https://github.com/waldeland/CNN-for-ASI/raw/master/F3/val/inline_405.png -O ${downdirval}/inline_405.png