IMPLEMENTASI ALGORITMA YOLOV7 DALAM SEGMENTASI CITRA JENIS BUAH APEL

Almarbai, Labibul Umam (2024) IMPLEMENTASI ALGORITMA YOLOV7 DALAM SEGMENTASI CITRA JENIS BUAH APEL. Other thesis, UPN Veteran Yogyakarta.

	Text ABSTRAK_123170092_Labibul Umam A.pdf Download (11kB)
	Text COVER_123170092_Labibul Umam A.pdf Download (162kB)
	Text DAFTAR ISI_123170092_Labibul Umam A-10-11.pdf Download (36kB)
	Text DAFTAR PUSTAKA_123170092_Labibul Umam A.pdf Download (146kB)
	Text LEMBAR PENGESAHAN_123170092_Labibul Umam A.pdf Download (5MB)
	Text SKRIPSI FULLTEXT_123170092_Labibul Umam A.pdf Restricted to Repository staff only Download (3MB)

Abstract

vi
ABSTRAK
Buah apel memiliki banyak jenis apel yang masing-masing memiliki rasa dan warna
yang bervariasi. Pada umumnya masyarakat menentukan jenis buah apel masih
menggunakan metode tradisional yaitu dengan cara mengidentifikasi tekstur, pola warna,
bau, dan ciri khas apel lainya, namun metode tradisional tersebut dapat mengakibatkan
persepsi yang berbeda saat mengidentifikasi buah apel oleh setiap orang karena sifat
subjektifnya. Seiring perkembangan zaman, teknologi pada berbagai bidang juga
dikembangkan, salah satunya adalah implementasi Computer Vision (Visi Komputer)
bidang Image Segmentation (Segmentasi Citra) pada makanan dan produk pertanian seperti
apel.
Penelitian ini menggunakan algoritma YOLOv7 dalam melakukan segmentasi citra
buah apel, karena pada tahun 2022 YOLOv7 mengungguli semua model deteksi objek
sebelumnya dalam hal kecepatan dan akurasi. Objek penelitian ini adalah delapan jenis buah
apel yaitu Braeburn, Crimson Snow, Golden, Golden Red, Granny Smith, Pink Lady, Red,
dan Red Delicious. Dilakukan transfer learning pada pre-trained model YOLOv7 dengan
dataset yang baru, agar model dapat melakukan segmentasi citra jenis-jenis buah apel dengan
baik. Model yang dihasilkan dari transfer learning dilakukan deploy menggunakan library
python streamlit sebagai user interface sistem.
Hasil penelitian ini menunjukkan bahwa model YOLOv7 yang sudah dilakukan
transfer learning dapat melakukan prediksi dengan sangat baik pada citra uji yang termasuk
kedalam dataset dan meghasilkan rata-rata precision sebesar 99,7% untuk semua kelas, rata-
rata recall sebesar 99,7% untuk semua kelas, dan rata-rata mAP sebesar 99,4% untuk semua
kelas. Namun performa model tersebut menurun ketika dilakukan segmentasi terhadap citra
buah apel diluar dataset, karena citra tersebut berbeda dengan citra dataset baik dari segi
kualitas maupun treatment yang dilakukan terhadap citra.
Kata kunci : Segmentasi apel, Implementasi YOLOv7, Streamlit
vii
ABSTRACT
There are many types of apples, each of which has varying flavors and colors. In
general, people still use traditional methods to determine the type of apple, namely by
identifying the texture, color pattern, smell and other characteristics of apples, but this
traditional method can result in different perceptions when identifying apples by each person
because of its subjective nature. As time goes by, technology in various fields is also being
developed, one of which is the implementation of Computer Vision in the field of Image
Segmentation in food and agricultural products such as apples.
This research uses the YOLOv7 algorithm to segment apple images, because in 2022
YOLOv7 will outperform all previous object detection models in terms of speed and
accuracy. The objects of this research are eight types of apples, namely Braeburn, Crimson
Snow, Golden, Golden Red, Granny Smith, Pink Lady, Red, and Red Delicious. Transfer
learning was carried out on the pre-trained YOLOv7 model with the new dataset, so that the
model could segment images of apple types well. The model resulting from transfer learning
is deployed using the Python Streamlit library as the system user interface.
The results of this research show that the YOLOv7 model that has carried out transfer
learning can make very good predictions on test images included in the dataset and produces
an average precision of 99.7% for all classes, an average recall of 99.7% for all classes,
and the average mAP is 99.4% for all classes. However, the model's performance decreases
when segmenting apple images outside the dataset, because these images are different from
the dataset images both in terms of quality and the treatment carried out on the images.
Keywords : Apple segmentation, YOLOv7 implementation, Streamlit

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Apple segmentation, YOLOv7 implementation, Streamlit
Subjects:	Z Bibliography. Library Science. Information Resources > ZA Information resources
Divisions:	Faculty of Engineering, Science and Mathematics > School of Engineering Sciences
Depositing User:	Eko Yuli
Date Deposited:	24 Jul 2024 03:46
Last Modified:	24 Jul 2024 03:46
URI:	http://eprints.upnyk.ac.id/id/eprint/40384

Actions (login required)

View Item