Show simple item record

dc.contributor.advisorMarcelita, Faldiena
dc.contributor.authorislam, Fatahillah saiful
dc.date.accessioned2026-08-12T08:42:15Z
dc.date.available2026-08-12T08:42:15Z
dc.date.issued2026
dc.identifier.urihttp://repository.ipb.ac.id/handle/123456789/178467
dc.description.abstractPenelitian ini membangun prototipe sistem keamanan parkir berbasis visi komputer dengan ESP32-CAM sebagai sumber video, YOLOv8s sebagai model deteksi kendaraan roda dua, server sebagai pusat pemrosesan, dashboard web sebagai antarmuka, serta Telegram Bot sebagai kanal notifikasi. Dataset terdiri atas 6.678 citra satu kelas Motorcycle yang dibagi menjadi 5.480 citra latih, 813 citra validasi, dan 385 citra uji. Model dilatih dari bobot YOLOv8s.pt dengan ukuran input 960 piksel dan batch size 24. Evaluasi validation set pada epoch ke-65 menghasilkan Precision 95,55%, Recall 83,11%, F1-Score 88,89%, mAP@0.5 sebesar 92,03%, dan mAP@0.5:0.95 sebesar 67,52%. Sistem mampu menerima stream kamera, mendeteksi dan melacak kendaraan, memeriksa posisi terhadap Region of Interest, mencatat event, menampilkan riwayat, mengekspor data, serta mengirimkan notifikasi missing dan removed. Integrasi sistem telah berjalan sebagai prototipe pemantauan parkir, dengan keterbatasan pada cakupan verifikasi lapangan dan kondisi pencahayaan rendah.
dc.description.abstractThis study developed a computer-vision-based parking security prototype using an ESP32-CAM as the video source, YOLOv8s as the motorcycle detection model, a server as the processing center, a web dashboard as the interface, and a Telegram Bot as the notification channel. The dataset contained 6,678 images of one Motorcycle class, consisting of 5,480 training, 813 validation, and 385 test images. The model was trained from YOLOv8s.pt weights with a 960-pixel input and a batch size of 24. Validation at epoch 65 yielded 95.55% precision, 83.11% recall, an F1-score of 88.89%, mAP@0.5 of 92.03%, and mAP@0.5:0.95 of 67.52%. The system received the camera stream, detected and tracked motorcycles, checked a Region of Interest, recorded events, displayed histories, exported data, and sent missing and removed notifications. The integration functioned as a parking monitoring prototype, with limitations related to field verification coverage and low-light conditions.
dc.description.sponsorship
dc.language.isoid
dc.publisherIPB Universityid
dc.titlePembangunan Sistem Keamanan Parkir Berbasis YOLOv8 Melalui Integrasi ESP32-CAM dan Notifikasi Telegramid
dc.title.alternative
dc.typeTugas Akhir
dc.subject.keywordESP32-CAMid
dc.subject.keywordparking securityid
dc.subject.keywordreal-time notificationid
dc.subject.keywordTelegram Botid
dc.subject.keywordYOLOv8id
dc.subtypeUndergraduate Theses


Files in this item

Thumbnail
Thumbnail
Thumbnail

This item appears in the following Collection(s)

Show simple item record