Dealing with Topological Information within a Fully Convolutional Neural Network

Etienne Decencière (1) , Santiago Velasco-Forero (1) , Fu Min , Juanjuan Chen , Hélène Burdin , Gervais Gauthier (2) , Bruno Lay (3) , Thomas Bornschloegl , Thérèse Baldeweck (4)

1 CMM - Centre de Morphologie Mathématique
2 ADCIS - Société ADCIS
3 ADCIS - Société ADCIS
4 L'Oréal recherche et innovation

A fully convolutional neural network has a receptive field of limited size and therefore cannot exploit global information, such as topological information. A solution is proposed in this paper to solve this problem, based on pre-processing with a geodesic operator. It is applied to the segmentation of histological images of pigmented reconstructed epidermis acquired via Whole Slide Imaging.

Mots clés: geodesic operators; Convolutional neural network; Mathematical morphology; Histological image segmentation

Domaines: Informatique [cs]; Vision par ordinateur et reconnaissance de formes [cs.CV]; Informatique [cs]; Traitement des images [eess.IV]; Sciences du Vivant [q-bio]; Médecine humaine et pathologie; Anatomie, Histologie, Anatomopathologie [q-bio.TO]; Statistiques [stat]; Apprentissage Machine

Etienne Decencière, Santiago Velasco-Forero, Fu Min, Juanjuan Chen, Hélène Burdin, et al.. Dealing with Topological Information within a Fully Convolutional Neural Network. Advanced Concepts for Intelligent Vision Systems (ACIVS), Sep 2018, Poitiers, France. ⟨hal-01877624⟩