Marichal Henry ; Randall Gregory

Abstract

This work presents the INBD network proposed by Gillert et al. in CVPR-2023and studies its application for delineating tree rings in RGB images of Pinustaeda cross sections captured by a smartphone (UruDendro dataset), which areimages with different characteristics from the ones used to train the method.The INBD network operates in two stages: first, it segments the background,pith, and ring boundaries. In the second stage, the image is transformed intopolar coordinates, and ring boundaries are iteratively segmented from the pithto the bark. Both stages are based on the U-Net architecture. The methodachieves an F-Score of 77.5, a mAR of 0.540, and an ARAND of 0.205 on theevaluation set. The code for the experiments is available athttps://github.com/hmarichal93/mlbrief_inbd.