diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000..5379447
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1,6 @@
+.ipynb_checkpoints
+*.ipynb
+.idea
+.vscode
+eurosat
+eurosat.tar.gz
diff --git a/README.md b/README.md
new file mode 100644
index 0000000..0a19b00
--- /dev/null
+++ b/README.md
@@ -0,0 +1,24 @@
+# Google Colab & Deep Learning VM
+
+## Google Deep Learning VM
+
+```bash
+export IMAGE_FAMILY="pytorch-latest-gpu"
+export ZONE="europe-west1-d"
+export INSTANCE_NAME="fch-tds"
+
+gcloud compute instances create $INSTANCE_NAME \
+  --zone=$ZONE \
+  --image-family=$IMAGE_FAMILY \
+  --image-project=deeplearning-platform-release \
+  --maintenance-policy=TERMINATE \
+  --accelerator="type=nvidia-tesla-p100,count=1" \
+  --metadata="install-nvidia-driver=True" \
+  --boot-disk-size=120GB \
+  --preemptible
+```
+https://github.com/gclouduniverse/gcp-notebook-executor
+https://github.com/gclouduniverse/nova-jupyterlab-extensions
+
+# How to launch from the CLI & get jupyterlab
+https://towardsdatascience.com/how-to-use-jupyter-on-a-google-cloud-vm-5ba1b473f4c2
\ No newline at end of file
diff --git a/google-colab-demo-pytorch.py b/google-colab-demo-pytorch.py
new file mode 100644
index 0000000..0795536
--- /dev/null
+++ b/google-colab-demo-pytorch.py
@@ -0,0 +1,339 @@
+# ---
+# jupyter:
+#   jupytext:
+#     formats: ipynb,py:percent
+#     text_representation:
+#       extension: .py
+#       format_name: percent
+#       format_version: '1.2'
+#       jupytext_version: 1.1.6
+#   kernelspec:
+#     display_name: Python 3
+#     language: python
+#     name: python3
+# ---
+
+# %% [markdown]
+# # Training a basic CNN with Ignite on Google Colaboratory
+# Toulouse Data Science 18/06/2019
+
+# %% [markdown]
+# ## Installation of custom dependencies
+
+# %%
+# Houston we have GPU
+# !nvcc -V
+# !nvidia-smi
+
+# %%
+# List existing dependencies
+# !pip list
+
+# %%
+# installation of custom dependencies
+# !pip3 install tqdm pytorch-ignite
+
+# %% [markdown]
+# ## Download data from google drive
+
+# %%
+# We use the mount as storage option but there are many
+# See https://colab.research.google.com/notebooks/io.ipynb#scrollTo=RWSJpsyKqHjH for reference
+try:
+    from google.colab import drive
+    drive.mount('/content/gdrive')
+except ImportError:
+    pass
+
+# %%
+# !cp /content/gdrive/My\ Drive/eurosat.tar.gz .
+# %%
+# !tar -zxf eurosat.tar.gz
+
+# %%
+# !ls eurosat/
+# !ls eurosat/train/
+
+# %%
+# various imports
+import glob
+import os
+import random
+from itertools import cycle
+
+import numpy as np
+import torch
+import torch.nn.functional as F
+from PIL import Image
+import ignite.engine
+import ignite.metrics
+from matplotlib import pyplot as plt
+from sklearn.metrics import average_precision_score
+from sklearn.metrics import precision_recall_curve
+from torch import nn
+from torch.optim import SGD
+from torch.utils.data import DataLoader, Dataset
+from torchvision.transforms import Compose, ToTensor, RandomHorizontalFlip, RandomVerticalFlip
+from tqdm import tqdm
+
+# %% [markdown]
+# ## Configuration
+
+# %%
+random.seed(2019)
+BATCH_SIZE = 64
+LEARNING_RATE = 0.01
+MOMENTUM = 0.9
+MAX_EPOCHS = 5
+
+# %% [markdown]
+# ## Datasets and dataloaders definitions
+
+
+# %%
+# Define datasets
+class EurosatDataset(Dataset):
+    def __init__(self, images_files, shuffle=True):
+        self.images = images_files
+        self.classes = list(set([f.split("/")[-2] for f in self.images]))
+        self.classes = sorted(self.classes)
+
+        if shuffle:
+            random.shuffle(self.images)
+
+    @classmethod
+    def from_fold(cls, data_dir, fold, shuffle=True):
+        images = glob.glob(os.path.join(data_dir, fold, "**", "*.jpg"))
+        return cls(images, shuffle=shuffle)
+
+    def split(self, n=0.75):
+        n = int(n * len(self))
+        return EurosatDataset(self.images[:n], shuffle=True), EurosatDataset(self.images[n:], shuffle=True)
+
+    def __getitem__(self, item):
+        img = self.images[item]
+        img = Image.open(img)
+        cls = self.classes.index(self.images[item].split("/")[-2])
+
+        return img, cls
+
+    def __len__(self):
+        return len(self.images)
+
+
+class TransformedDataset(Dataset):
+    def __init__(self, dataset, transform=None, target_transform=None):
+        self.dataset = dataset
+        self.transform = transform
+        self.target_transform = target_transform
+
+    def __len__(self):
+        return len(self.dataset)
+
+    def __getitem__(self, item):
+        x, y = self.dataset[item]
+        if self.transform is not None:
+            x = self.transform(x)
+        if self.target_transform is not None:
+            y = self.target_transform(y)
+
+        return x, y
+
+
+# %%
+# Get data loaders functions
+
+def get_datasets(root_dir):
+
+    train_data_transform = Compose([RandomHorizontalFlip(), RandomVerticalFlip(), ToTensor()])
+    val_data_transform = Compose([ToTensor()])
+
+    train_dataset = EurosatDataset.from_fold(data_dir=root_dir, fold="train")
+    train_dataset, val_dataset = train_dataset.split(n=0.7)
+    test_dataset = EurosatDataset.from_fold(data_dir=root_dir, fold="test")
+
+    train_dataset = TransformedDataset(dataset=train_dataset, transform=train_data_transform)
+    val_dataset = TransformedDataset(dataset=val_dataset, transform=val_data_transform)
+    test_dataset = TransformedDataset(dataset=test_dataset, transform=val_data_transform)
+
+    return train_dataset, val_dataset, test_dataset
+
+
+def get_data_loaders(train_dataset, val_dataset):
+
+    train_loader = DataLoader(dataset=train_dataset, shuffle=True, batch_size=BATCH_SIZE)
+
+    val_loader = DataLoader(dataset=val_dataset, shuffle=False, batch_size=BATCH_SIZE)
+
+    return train_loader, val_loader
+
+
+# %% [markdown]
+# ## Model definition
+
+
+# %%
+# Define model to be trained
+class SimpleCNN(nn.Module):
+    def __init__(self):
+        super(SimpleCNN, self).__init__()
+        self.conv1 = nn.Sequential(nn.Conv2d(3, 32, kernel_size=3, padding=1), nn.ReLU())
+        self.conv2 = nn.Sequential(nn.Conv2d(32, 64, kernel_size=3, padding=1), nn.ReLU())
+        self.conv3 = nn.Sequential(nn.Conv2d(64, 64, kernel_size=3, padding=1), nn.ReLU())
+        self.conv4 = nn.Sequential(nn.Conv2d(64, 96, kernel_size=3, padding=1), nn.ReLU())
+        self.fc1 = nn.Sequential(nn.Linear(4 * 4 * 96, 32), nn.ReLU())
+        self.fc2 = nn.Linear(32, 10)
+
+    def forward(self, x):
+        # 64x64x3
+        x = F.max_pool2d(self.conv1(x), 2)  # 32x32x32
+        x = F.max_pool2d(self.conv2(x), 2)  # 16x16x64
+        x = F.max_pool2d(self.conv3(x), 2)  # 8x8x64
+        x = F.max_pool2d(self.conv4(x), 2)  # 4x4x96
+        x = F.dropout2d(x, training=self.training)
+        x = x.view(-1, 4 * 4 * 96)
+        x = self.fc1(x)
+        x = F.dropout(x, training=self.training)
+        x = self.fc2(x)
+        return F.log_softmax(x, dim=-1)
+
+
+# %% [markdown]
+# ## Training initialization
+
+# %%
+# Initialize datasets and models
+train_dataset, val_dataset, test_dataset = get_datasets(root_dir="./eurosat")
+train_loader, val_loader = get_data_loaders(train_dataset, val_dataset)
+model = SimpleCNN()
+
+device = 'cpu'
+
+if torch.cuda.is_available():
+    device = 'cuda'
+
+# %%
+# initialize training resources
+optimizer = SGD(model.parameters(), lr=LEARNING_RATE, momentum=MOMENTUM, nesterov=True)
+trainer = ignite.engine.create_supervised_trainer(model, optimizer, F.nll_loss, device=device)
+evaluator = ignite.engine.create_supervised_evaluator(
+    model, metrics={
+        'accuracy': ignite.metrics.Accuracy(),
+        'nll': ignite.metrics.Loss(F.nll_loss)
+    }, device=device)
+
+desc = "ITERATION - loss: {:.2f}"
+pbar = tqdm(initial=0, leave=False, total=len(train_loader), desc=desc.format(0))
+
+
+@trainer.on(ignite.engine.Events.ITERATION_COMPLETED)
+def log_training_loss(engine):
+    iter = (engine.state.iteration - 1) % len(train_loader) + 1
+
+    if iter % 100 == 0:
+        pbar.desc = desc.format(engine.state.output)
+        pbar.update(100)
+
+
+@trainer.on(ignite.engine.Events.EPOCH_COMPLETED)
+def log_training_results(engine):
+    pbar.refresh()
+    evaluator.run(train_loader)
+    metrics = evaluator.state.metrics
+    avg_accuracy = metrics['accuracy']
+    avg_nll = metrics['nll']
+    tqdm.write("Training Results - Epoch: {}  Avg accuracy: {:.2f} Avg loss: {:.2f}".format(
+        engine.state.epoch, avg_accuracy, avg_nll))
+
+
+@trainer.on(ignite.engine.Events.EPOCH_COMPLETED)
+def log_validation_results(engine):
+    evaluator.run(val_loader)
+    metrics = evaluator.state.metrics
+    avg_accuracy = metrics['accuracy']
+    avg_nll = metrics['nll']
+    tqdm.write("Validation Results - Epoch: {}  Avg accuracy: {:.2f} Avg loss: {:.2f}".format(
+        engine.state.epoch, avg_accuracy, avg_nll))
+
+    pbar.n = pbar.last_print_n = 0
+
+
+# %% [markdown]
+# ## Run training
+
+# %%
+# go !
+trainer.run(train_loader, max_epochs=MAX_EPOCHS)
+pbar.close()
+
+# %% [markdown]
+# ## Evaluate on test dataset (PR curve)
+
+# %%
+test_loader = DataLoader(dataset=test_dataset, shuffle=False, batch_size=BATCH_SIZE)
+
+# %%
+test_iterator = iter(test_loader)
+
+# %%
+all_y_preds = []
+all_y_trues = []
+for x, y_true in test_iterator:
+    if torch.cuda.is_available():
+        y_pred = F.softmax(model(x.cuda()), dim=-1).detach().cpu().numpy()
+        y_true = y_true.cpu().numpy()
+    else:
+        y_pred = F.softmax(model(x), dim=-1).detach().numpy()
+        y_true = y_true.numpy()
+    y_true = np.eye(10)[y_true]
+    all_y_trues.extend(y_true)
+    all_y_preds.extend(y_pred)
+y_pred = np.asarray(all_y_preds)
+y_true = np.asarray(all_y_trues)
+
+# %%
+# Compute precision and recall for each class
+precision = dict()
+recall = dict()
+average_precision = dict()
+for i in range(10):
+    precision[i], recall[i], _ = precision_recall_curve(y_true[:, i], y_pred[:, i])
+    average_precision[i] = average_precision_score(y_true[:, i], y_pred[:, i])
+
+# %%
+# A "micro-average": quantifying score on all classes jointly
+precision["micro"], recall["micro"], _ = precision_recall_curve(y_true.ravel(), y_pred.ravel())
+average_precision["micro"] = average_precision_score(y_true, y_pred, average="micro")
+print('Average precision score, micro-averaged over all classes: {0:0.2f}'.format(average_precision["micro"]))
+
+# %%
+# Plot pr curve (code taken from sklearn)
+colors = cycle(['navy', 'turquoise', 'darkorange', 'cornflowerblue', 'teal', 'red', 'yellow', 'purple', 'green'])
+plt.figure(figsize=(7, 8))
+f_scores = np.linspace(0.2, 0.8, num=4)
+lines = []
+labels = []
+for f_score in f_scores:
+    x = np.linspace(0.01, 1)
+    y = f_score * x / (2 * x - f_score)
+    l, = plt.plot(x[y >= 0], y[y >= 0], color='gray', alpha=0.2)
+    plt.annotate('f1={0:0.1f}'.format(f_score), xy=(0.9, y[45] + 0.02))
+lines.append(l)
+labels.append('iso-f1 curves')
+l, = plt.plot(recall["micro"], precision["micro"], color='gold', lw=2)
+lines.append(l)
+labels.append('micro-average Precision-recall (area = {0:0.2f})' ''.format(average_precision["micro"]))
+for i, color in zip(range(10), colors):
+    l, = plt.plot(recall[i], precision[i], color=color, lw=2)
+    lines.append(l)
+    labels.append('Precision-recall for class {0} (area = {1:0.2f})'
+                  ''.format(test_dataset.dataset.classes[i], average_precision[i]))
+fig = plt.gcf()
+fig.subplots_adjust(bottom=0.25)
+plt.xlim([0.0, 1.0])
+plt.ylim([0.0, 1.05])
+plt.xlabel('Recall')
+plt.ylabel('Precision')
+plt.title('Extension of Precision-Recall curve to multi-class')
+plt.legend(lines, labels, loc=(0, -1.0), prop=dict(size=14))
+plt.show()
diff --git a/requirements.txt b/requirements.txt
new file mode 100644
index 0000000..e69de29
diff --git a/train_test_split.py b/train_test_split.py
new file mode 100644
index 0000000..c6e1684
--- /dev/null
+++ b/train_test_split.py
@@ -0,0 +1,42 @@
+import glob
+import os
+import random
+import shutil
+from collections import defaultdict
+import tqdm
+
+random.seed(2019)
+# %%
+all_images = glob.glob(os.path.join("data", "raw", "**", "*.jpg"))
+
+per_class = defaultdict(list)
+
+for f in all_images:
+    cls = f.split("/")[-2].lower()
+    per_class[cls].append(f)
+
+# %%
+train_dir = "./data/train/"
+test_dir = "./data/test/"
+
+shutil.rmtree(train_dir, ignore_errors=True)
+shutil.rmtree(test_dir, ignore_errors=True)
+os.makedirs(train_dir, exist_ok=True)
+os.makedirs(test_dir, exist_ok=True)
+
+for cls in tqdm.tqdm(per_class):
+    list_images = per_class[cls]
+    random.shuffle(list_images)
+
+    split = int(0.75 * len(list_images))
+
+    train_images = list_images[:split]
+    test_images = list_images[split:]
+    os.makedirs(os.path.join(train_dir, cls), exist_ok=True)
+    os.makedirs(os.path.join(test_dir, cls), exist_ok=True)
+
+    for i, f in enumerate(train_images):
+        shutil.copyfile(f, os.path.join(train_dir, cls, "{}_{}.jpg".format(cls, i)))
+    for i, f in enumerate(test_images):
+        shutil.copyfile(f, os.path.join(test_dir, cls, "{}_{}.jpg".format(cls, i)))
+