Deteksi Hewan di Taman Safari Prigen Menggunakan Mask R-CNN

Jahja, Yasril (2022) Deteksi Hewan di Taman Safari Prigen Menggunakan Mask R-CNN. Other thesis, Institut Teknologi SepuluhNopember.

Text 07111840000076-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 April 2025. Download (7MB) | Request a copy

Abstract

Object Detection adalah salah satu teknologi yang berkembang pada saat ini. Banyak peneliti yang tertarik dengan pengembangan algoritma yang ada untuk mengoptimalkan kinerja pendeteksian objek pada suatu citra. Sistem deteksi objek dapat digunakan dalam membantu dan mempermudah pekerjaan manusia. Salah satunya dalam mendeteksi objek pada mobil otonom. Pendeteksian objek pada mobil otonom dibantu oleh Beberapa Sensor, seperti kamera, lidar, dan radar.Sensor ini berperan penting dalam hal navigasi dan pengenalan lingkungan sekitar. Salah satu objek yang menjadi halangan di jalan adalah hewan. Hewan-hewan yang lalu-lalang pada lajur mobil pengunjung Taman Safari Prigen susah untuk diidentifikasi karena hewan tersebut tidak sesuai tempat lingkungan yang berada. Tugas akhir ini mencoba mendeteksi hewan dengan menggunakan Mask R-CNN. Data yang digunakan untuk training merupakan gambar dari 5 jenis hewan, yaitu llama rusa, singa, unta, dan zebra yang berasal dari website penyedia dataset. Dataset ini dilabel secara manual dan bentuk labelnya adalah polygon. Deteksi objek berupa hewan yang dihasilkan adalah instance segmentation dan bounding box dari jenis hewan tersebut. Model yang telah dilatih kemudian di evaluasi menggunakan dataset testing agar mendapatkan F1-Score. F1-Score terbaik dengan nilai 0.30 untuk epoch 50. Kemudian model diuji dengan footage video pada jalur safari adventure di Taman Safari Prigen. Model diuji berdasarkan tingkat keberhasilannya jenis hewan, pengaruh intensitas cahaya, dan sudut pandang hewan. Dari lima jenis hewan yang diuji, unta dan zebra memiliki tingkat keberhasilan yang tinggi dan rusa menjadi hewan dengan tingkat keberhasilan yang paling rendah.
=================================================================================================================================
The object detection is one of the technologies presently being developed. There are many researchers are interested to optimize the object detection performance in an image by developing the existing algorithms. The object detection systems can be used to assist and facilitate human work. One of the uses is in detecting objects in autonomous cars. The object detection in autonomous cars is assisted by several sensors, such as cameras, lidars and radars. This sensor plays an important role in terms of navigation and recognition of the surrounding environment. One of the objects that become hindrances in the way is animal. Animals passing by in the car lanes of visitors to the Prigen Safari Park are difficult to identify because these animals do not fit in the environment where they should be. This final project attempts to detect the animals using the R-CNN Mask. The data used for training are images of five types of animals, namely llamas, deer, lions, camels and zebras, which come from the dataset provider's website. This dataset is labeled manually and the shape of the label is polygon. The result of the object detection with the form of animal is the instance segmentation and bounding box of that type of animal. The model that has been trained is then evaluated using dataset testing in order to get the F1 Score. The best F1 score with a value of 0.30 for epoch 50. Then the model was tested with video footage on the safari adventure track at Prigen Safari Park. The model was tested based on the success rate of the type of animal, the effect of light intensity, and the animal's point of view. Of the five types of animals tested, camels and zebras have the highest success rate and deer are the animals with the lowest success rate.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Mask R-CNN, Instance segmentation, deteksi hewan, Mask R-CNN, Instance segmentation, animal Detection
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7871.674 Detectors. Sensors
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20201-(S1) Undergraduate Thesis
Depositing User:	Yasril Jahja
Date Deposited:	02 Feb 2023 14:55
Last Modified:	02 Feb 2023 14:55
URI:	http://repository.its.ac.id/id/eprint/96033

Actions (login required)

View Item