Tutorial: Introduction to convolutional neural networks (CNNs)

This tutorial introduces a simple CNN with keras and tensorflow to classify remote sensing images from the UC Merced dataset.

Getting Started

1. Clone this repository to your local machine. In your terminal type:

git clone https://github.com/langnico/DL_tutorial_RS.git

2. Download the data and pre-trained models from this link:

Google Drive

Move the directories into the code directory DL_tutorial_RS/. The directory tree should look like this:

• DL_tutorial_RS/
  • data/
  • pretrained_model_simpleCNN/

Prerequisites

We are going to execute the code in a jupyter notebook and use keras with a tensorflow backend.

Therefore, we need to install:

• python3
• jupyter
• tensorflow

Further we will need the python packages/modules:

• sklearn
• numpy
• matplotlib
• keras

Installing

We propose to install python via anaconda.

1. Install Anaconda and read the Anaconda tutorial (20min)

2. Create a new environment: conda create --name DLenv python=3.6

3. Activate the new conda environment (for conda 4.6 and later versions)

   • Windows: conda activate DLenv
   • Linux and macOS: conda activate DLenv

   For versions prior to conda 4.6, use:

   • Windows: activate DLenv
   • Linux, macOS: source activate DLenv

   --> now your terminal prompt should start with (DLenv)

4. Install the following packages in your activated DLenv:

   conda install jupyter
   conda install scikit-learn
   conda install matplotlib
   conda install keras
   

5. Install tensorflow with pip in the activated anaconda environment (DLenv).

   For example install the current stable release for CPU-only:

   pip install tensorflow
   

   Alternatively, install tensorflow with GPU support using: tensorflow-gpu. For the GPU version you might have to follow the official tensorflow installation instructions or check the anaconda installation instructions.

Verify your installation

1. In the activated DLenv type which jupyter. This should point to the python installation in your conda env e.g. /username/anaconda3/envs/DLenv/bin/jupyter

2. Open a terminal and go to the location of the file: installation_check.ipynb

   Then open this jupyter notebook with: jupyter notebook installation_check.ipynb

   NOTE: If this does not automatically open a browser showing the notebook, then open a browser (Firefox, Chrome) and type: http://localhost:8888/notebooks/installation_check.ipynb

   Then select the first cell containing the imports and click on the > Run Button. If your installation was successful, the output should be something like this:

   Using TensorFlow backend.
   successfully imported
   keras version:  2.2.4
   

Code inspirations

• https://github.com/tgjeon/Keras-Tutorials
• https://github.com/flyyufelix/cnn_finetune

Authors

• Nico Lang