Open AccessSegmentação de imagens de ultrassonografia da carcaça em pequenos ruminantes utilizando Deep Learning
Author(s) -
Tiago Oliveira,
José Lindenberg Rocha Sarmento,
Luiz A. S. Figueiredo Filho,
Romuere Silva
Publication year - 2021
Language(s) - English
Resource type - Conference proceedings
DOI - 10.5753/enucompi.2021.17765
Subject(s) - context (archaeology) , agribusiness , segmentation , convolutional neural network , anatomy , computer science , geography , fasciculation , artificial intelligence , cartography , agricultural science , biology , agriculture , archaeology
Brazil is one of the main countries operating in the agribusiness sector. Sheep and goat farming is one of the segments of Brazilian agribusiness. The evaluation of each carcass of goats and sheep is carried out by a specialist who evaluates them based on visual aspects, being susceptible to errors in the final evaluation. In this context, the purpose of this work is to use Convolutional Neural Networks to segment the Longissimus dorsi muscle area in ultrasonographic images of small ruminants. Our experiments showed that the PSPNet CNN architecture achieved the results with an Intersect over Union (IoU) rate of 0.89. It was possible to obtain a precise segmentation of the images, which will allow the producer to correctly diagnose the measurements of the animals with greater practicality and saving time.