Semantic Segmentation-based Automatic Wood Cell Detection from Raster Images

A comprehensive examination of all of the wood’s elements is required in order to identify wood by its characteristics. There is no comprehensive solution to this important issue in structural wood research. For most tree species, the vessels of trees and their mutual location have a distinct personality. Semantic segmentation has become increasingly popular in a variety of scientific fields in recent years. Automated image annotation is a critical step in extracting functional information. This study focuses on the development of a method for automatic tree vessel detection from raster images using semantic segmentation techniques. We present three model architectures and compare them to find the optimum effectiveness of the segmentation. Segmentation involves extracting semantic content in the neural network encoder and then adding/replacing it in the decoder.