PALM OIL POLLEN DETECTION AND COUNTING SYSTEM USING YOLO V7 ALGORITHM

Faroza, Mohammad Rifqi Nur (2024) PALM OIL POLLEN DETECTION AND COUNTING SYSTEM USING YOLO V7 ALGORITHM. Undergraduate thesis, Universitas Muhammadiyah Malang.

[thumbnail of PENDAHULUAN.pdf]
Preview
Text
PENDAHULUAN.pdf

Download (1MB) | Preview
[thumbnail of BAB I.pdf]
Preview
Text
BAB I.pdf

Download (295kB) | Preview
[thumbnail of BAB II.pdf]
Preview
Text
BAB II.pdf

Download (3MB) | Preview
[thumbnail of BAB III.pdf]
Preview
Text
BAB III.pdf

Download (2MB) | Preview
[thumbnail of BAB IV.pdf] Text
BAB IV.pdf
Restricted to Registered users only

Download (5MB) | Request a copy
[thumbnail of BAB V.pdf] Text
BAB V.pdf
Restricted to Registered users only

Download (267kB) | Request a copy
[thumbnail of LAMPIRAN.pdf]
Preview
Text
LAMPIRAN.pdf

Download (2MB) | Preview
[thumbnail of Mohammad Rifqi Nur Faroza_201910130311014_POSTER.pdf] Text
Mohammad Rifqi Nur Faroza_201910130311014_POSTER.pdf
Restricted to Registered users only

Download (7MB) | Request a copy

Abstract

For sustainability and efficiency in observing the percentage of oil palm pollen germination that is still observed and calculated manually, in this case it allows less effectiveness and efficiency of pollinators when determining the viability of good pollen for pollination, the length of the process of determining pollen viability and the occurrence of miscalculation bias. In this study, we evaluated the robustness of the training period for You Only Look Once (YOLO) v7x, a Convolutional Neural Network (CNN) model for developing live and dead pollen classification. Images were annotated using bounding boxes and trained with training rates of 400, 500, 600, 700, 800, 900 and 1000 epochs. To determine the optimal performance on the test set, the model was trained on multiple epochs, and training was stopped when the test performance (classification accuracy, precision, and recall) started to decrease. The results obtained show that the precision value of 1000 epochs has the highest weight value, with a precision value of 94%. As for the training time, the shortest time is 400 epochs with a model training time of 2.355 hours. For the most appropriate and efficient training results are the results of training using 900 epochs because the weight produced is quite high, namely 94% difference of 0.015 with the results of training 1000 epochs and the time when training is 0.368 hours faster than using 1000 epochs.

Item Type: Thesis (Undergraduate)
Student ID: 201910130311014
Keywords: Object Detection, YOLOv7, Artificial Intelegence, CNN
Subjects: T Technology > TK Electrical engineering. Electronics Nuclear engineering
Divisions: Faculty of Engineering > Department of Electrical Engineering (20201)
Depositing User: 201910130311014 farozakyalafasy
Date Deposited: 20 Jan 2024 02:08
Last Modified: 20 Jan 2024 02:08
URI: https://eprints.umm.ac.id/id/eprint/2746

Actions (login required)

View Item
View Item