Rotation-invariant NN for learning naturally un-oriented data

Abstract : Deep convolutional neural networks accuracy is heavily impacted by the rotations of the input data. In this paper, we propose a convolutional predictor that is invariant to rotations in the input. This architecture is capable of predicting the angular orientation without angle-annotated data. Furthermore, the predictor maps continuously the random rotation of the input to a circular space of the prediction. For this purpose, we use the roto-translation properties existing in the Scattering Transform Networks with a series of 3D Convolutions. We validate the results by training with upright and randomly rotated samples. This allows further applications of this work on fields like automatic re-orientation of randomly oriented datasets.
Type de document :
Pré-publication, Document de travail
42ème journée ISS France. 2019
Liste complète des métadonnées
Contributeur : Rosemberg Rodriguez Salas <>
Soumis le : lundi 18 février 2019 - 11:26:00
Dernière modification le : mardi 19 mars 2019 - 23:43:25


Fichiers produits par l'(les) auteur(s)


  • HAL Id : hal-02022802, version 1


Rosemberg Rodriguez Salas, Petr Dokládal, Eva Dokladalova. Rotation-invariant NN for learning naturally un-oriented data. 42ème journée ISS France. 2019. 〈hal-02022802〉



Consultations de la notice


Téléchargements de fichiers