Banca de QUALIFICAÇÃO: FRANCISCA ADRIANA FERREIRA DE ANDRADE

Uma banca de QUALIFICAÇÃO de MESTRADO foi cadastrada pelo programa.
STUDENT : FRANCISCA ADRIANA FERREIRA DE ANDRADE
DATE: 30/10/2021
TIME: 09:00
LOCAL: Sala do Google meet
TITLE:

Deep learning of machines: potential for use in predicting seed germination of Pityrocarpa moniliformis (Benth.) Luckow & R. W. Jobson

KEY WORDS:

catanduva, seed quality, deep learning, automation

PAGES: 61
BIG AREA: Ciências Agrárias
AREA: Recursos Florestais e Engenharia Florestal
SUBÁREA: Silvicultura
SPECIALTY: Sementes Florestais
SUMMARY:

The evaluation of seed germination through deep machine learning has shown potential as a complementary method in the analysis of seed quality. This work aims to evaluate the efficiency of using deep machine learning with convolutional neural networks to determine the quality of seeds of Pityrocarpa moniliformis (Benth.) Luckow & R. W. Jobson. 1000 seeds were randomly selected from 4 different lots, 250 for each. For seed image analysis, a scanner was used to obtain the images of each seed, which were used to implement, train and test the convolutional neural networks in the computational algorithm created, in order to compare the results obtained with traditional non-automated methods. 5 replicates of 50 seeds were submitted to germination, germination speed index, seedling length, fresh and dry mass tests. The experimental design used was completely randomized, with 5 replications. Data were subjected to analysis of variance, means compared by Scott-Knott test at 5% probability and the statistical program was Sisvar®. The pre-trained networks after 5 times of execution show a tendency to improve the accuracy, however it also shows an indication of overfitting, since the performance in the training data is better than in the validation data (test), the recall (sensitivity) was greater than 80% in all for germinates class. The recall value was much lower for the non-germinating class, both below 20%. For the customized model, 85% recall was obtained for the class that does not germinate and 18% for germinates, there may have been an overlap or inversion in the recognition of the classes. For comparisons between computerized analysis, the germination test and IVG, it is necessary to obtain more clarity and more representative data regarding the class that does not germinate. The results of the pre-trained networks and the customized model are very promising for both the training set and the test set, because it is possible to verify the viability of the catanduva germination, but the analyzes indicate the need to improve and adjust the pre-processing of the images and the data augmentation.

BANKING MEMBERS:
Presidente - 1880265 - MARCIO DIAS PEREIRA
Externa ao Programa - 2938129 - LAURA EMMANUELLA ALVES DOS SANTOS SANTANA DE OLIVEIRA
Externa à Instituição - RAQUEL MARIA DE OLIVEIRA PIRES