Analisis Model Ringan Pra-terlatih Deep Learning Untuk Klasifikasi Spesies Burung

Anatasya, Remet Tirzah (2026) Analisis Model Ringan Pra-terlatih Deep Learning Untuk Klasifikasi Spesies Burung. Bachelor thesis, Institut Teknologi Kalimantan.

	Text 11211072_abstract_en.pdf Restricted to Repository staff only until 26 April 2028. Download (191kB) | Request a copy
	Text 11211072_abstract_id.pdf Restricted to Registered users only until 26 April 2028. Download (237kB) | Request a copy
	Text 11211072_approval_sheet.pdf Restricted to Registered users only until 26 April 2028. Download (132kB) | Request a copy
	Text 11211072_bibliography.pdf Restricted to Registered users only until 26 April 2028. Download (263kB) | Request a copy
	Text 11211072_chapter_1.pdf Restricted to Repository staff only until 26 April 2028. Download (330kB) | Request a copy
	Text 11211072_chapter_2.pdf Restricted to Repository staff only until 26 April 2028. Download (1MB) | Request a copy
	Text 11211072_chapter_3.pdf Restricted to Repository staff only until 26 April 2028. Download (356kB) | Request a copy
	Text 11211072_chapter_4.pdf Restricted to Repository staff only until 26 April 2028. Download (1MB) | Request a copy
	Text 11211072_conclusions.pdf Restricted to Repository staff only until 26 April 2028. Download (266kB) | Request a copy
	Text 11211072_cover.pdf Restricted to Registered users only until 26 April 2028. Download (317kB) | Request a copy
	Text 11211072_form-ta020.pdf Restricted to Repository staff only until 26 April 2028. Download (129kB) | Request a copy
	Text 11211072_illustrations.pdf Restricted to Repository staff only until 26 April 2028. Download (237kB) | Request a copy
	Text 11211072_paper.pdf Restricted to Repository staff only until 26 April 2028. Download (843kB) | Request a copy
	Text 11211072_preface.pdf Restricted to Registered users only until 26 April 2028. Download (231kB) | Request a copy
	Text 11211072_publishing_agreement.pdf Restricted to Registered users only until 26 April 2028. Download (229kB) | Request a copy
	Text 11211072_table_of_content.pdf Restricted to Repository staff only until 26 April 2028. Download (276kB) | Request a copy
	Text 11211072_tables.pdf Restricted to Repository staff only until 26 April 2028. Download (199kB) | Request a copy

Abstract

Bird biodiversity plays a vital role in maintaining ecological balance but is highly vulnerable to urbanization and environmental degradation. Manual identification of bird species requires considerable time and expertise, making automated classification systems essential. This study develops an efficient bird species classification framework using lightweight deep learning models. Six pretrained architectures were evaluated: MobileViT-V1, MobileViT-V2, EfficientNetV2-B3, ResNet-18, MobileNetV3, and ShuffleNetV2. The dataset, obtained from TensorFlowDataset, consists of 200 bird species with a total of 11,788 images. The research procedure included data preprocessing, model training, and performance evaluation using accuracy, precision, recall, F1-score, and latency metrics. Results indicate that EfficientNetV2-B3 achieved the highest accuracy (63%) and F1-score (0.63). MobileViT-V2 provided a balanced trade-off with 57% accuracy and 1.833 ms latency. ResNet-18 and MobileNetV3 demonstrated low latency (<1 ms) with moderate accuracy (53% and 56%), making them suitable for real-time applications. ShuffleNetV2 recorded the lowest latency (0.220 ms) but poor accuracy (34%), while MobileViT-V1 showed low accuracy (42%) with relatively high latency (2.514 ms). Based on comparative analysis, MobileNetV3 is recommended as the optimal model for deployment on resource-constrained devices, offering the best balance between accuracy and computational efficiency

Item Type:	Thesis (Bachelor)
Subjects:	A General Works > AI Indexes (General)
Divisions:	Jurusan Matematika dan Teknologi Informasi > Informatika
Depositing User:	Remet Tirzah Anatasya
Date Deposited:	09 Jan 2026 03:27
Last Modified:	09 Jan 2026 03:27
URI:	http://repository.itk.ac.id/id/eprint/25022

Actions (login required)

View Item