Documentation for package did¶
This is the documentation for the did package, the implementation of paper Measuring dissimilarity with diffeomorphism invariance, by Théophile Cantelobre, Alessandro Rudi, Carlo Ciliberto and Benjamin Guedj.
The paper introduces a dissimilarity between data spaces (for example between images or time series) that is invariant by diffeomorphisms. Practically, the dissimilarity between an image and a rotated version of the same image is close to zero with the dissimilarity we introduce.
This library is a campanion for the paper. The source code is available at LINK, along with sample data and scripts (allowing to reproduce the results presented in the paper).
To get started with did (installation and first examples), checkout the Getting started page below.