Here you can find some MATLAB codes used in our publications and researches. This page will be updated as soon as the related publications are available. The codes are provided in a compressed file in which a main file calls the functions (if any) in the subfolders. Simply extract the files to a folder and 'cd' it in MATLAB (set as current directory). Once this is done, simply run the main file. Enjoy!

Our work is shared under a creative common license! Please check it!

  • Dense semantic labeling of sub-decimeter resolution images with convolutional neural networks [citation] [project page]
Here we provide trained networks, code to do prediction and code to fine tune them (MATLAB). Note that these are minimum working examples, and results might slightly differ from those reported in the paper (but are basically the same). In order to use these codes you will need to install and compile MatConvNet (tested on 1.0-beta19) and cropRectanglesMex. These are independently maintained projects so don't ask me to redistribute their code. Have fun!

  • Spectral alignment of multi-temporal cross-sensor images with automated kernel correlation analysis [citation] [project page]

The MATLAB code implements the fully automatic cross-sensor relative spectral alignment technique presented in the paper. It relies on the automatic model selection for the kernel canonical correlation analysis transformation to find directions of cross-correlation between the data. Change detection is then evaluated using the change vector analysis. For more details, please see the paper. For code and change detection data, see project page link.

This toolbox provides a working example of the code used for the paper experiments. Note that the site also provides the Pines Image to readily test it off-the-shelf!
We provide a simple MATLAB implementation of the semisupervised manifold alignment for the paper. This is a MWE, and you may have to adapt it to your needs.

  • Unsupervised change detection with kernels [citation] [MATLAB code: zip]:

This example implements the automatic change detection algorithm presented at SPIE 2010, IGARSS 2011, and in the paper 'Unsupervised change detection with kernels'. It consists in three steps: initialization (histogram-based), automatic parameter tuning and change map generation, with classical kmeans, gaussian kernel kmeans and by clustering the difference image in the feature spaces. Loops allows to perform different experiments and the user can choose different parameters (number of experiments, kernel function and parameters to search, number of pseudo training samples,...). Moreover, one can easily adapt the code for comparisons to other algorithms and images, as well as adapting the code for personal developments. Thanks to Frank de Morsier for some bug fixes!

This toolbox groups many AL techniqes (in particular those reviewed in the 2011 JSTSP paper). It is implemented in MATLAB, also thanks to the TORCH 3 machine learning library. It has been coded by Jordi and Devis, give credit!