Skip to Main content Skip to Navigation
Conference papers

Semantic image segmentation based on spatial relationships and inexact graph matching

Abstract : We propose a method for semantic image segmentation, combining a deep neural network and spatial relationships between image regions, encoded in a graph representation of the scene. Our proposal is based on inexact graph matching, formulated as a quadratic assignment problem applied to the output of the neural network. The proposed method is evaluated on a public dataset used for segmentation of images of faces and compared to the U-Net deep neural network that is widely used for semantic segmentation. Preliminary results show that our approach is promising. In terms of Intersection-over-Union of region bounding boxes, the improvement is of 2.4% in average, compared to U-Net, and up to 24.4% for some regions. Further improvements are observed when reducing the size of the training dataset (up to 8.5% in average).
Complete list of metadata
Contributor : Isabelle Bloch <>
Submitted on : Wednesday, December 16, 2020 - 8:47:41 PM
Last modification on : Tuesday, March 30, 2021 - 11:33:46 AM
Long-term archiving on: : Wednesday, March 17, 2021 - 8:12:28 PM


Files produced by the author(s)



Jérémy Chopin, Jean-Baptiste Fasquel, Harold Mouchère, R. Dayot, I. Bloch. Semantic image segmentation based on spatial relationships and inexact graph matching. International Conference on Image Processing Theory, Tools and Applications (IPTA), Nov 2020, Paris, France. ⟨10.1109/IPTA50016.2020.9286611⟩. ⟨hal-02916165⟩



Record views


Files downloads