Article Per Year (5 Year)
p-Index From 2020 - 2025
5.678
P-Index
This Author published in this journals
All Journal IAES International Journal of Artificial Intelligence (IJ-AI) Transmisi: Jurnal Ilmiah Teknik Elektro Bulletin of Electrical Engineering and Informatics Bulletin of Electrical Engineering and Informatics Jurnal Mahasiswa TEUB Jurnal EECCIS Bulletin of Electrical Engineering and Informatics Network Engineering Research Operation [NERO] Indonesian Journal of Electrical Engineering and Computer Science Jurnal INFOTEL
Mudjirahardjo, Panca
Jurusan Teknik Elektro Fakultas Teknik Universitas Brawijaya
Computer Science & IT Education Electrical & Electronics Engineering Engineering

Published : 78 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 78 Documents
Search

 4   5   6   7   8 
SISTEM PENGHITUNG KENDARAAN DENGAN OPENCV DAN MODEL PEDETEKSI PRA-TERLATIH MOBILENET SSD Fauzi, Maher; Mudjirahardjo, Panca; Razak, Angger Abdul
Jurnal Mahasiswa TEUB Vol. 12 No. 1 (2024)
Publisher : Jurnal Mahasiswa TEUB

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Accurate vehicle detection and tracking is required in visual-based systems to perform counting, therefore vehicle detection by utilizing artificial intelligence networks to process image data CNN (Convolutional Neural Network) is introduced, but in implementing CNN several problems arise, CNN models that are often used are better in accuracy and precision, but often use excessive processing power and computing resources to work optimally, they are not compatible with low-budget platforms and systems, and not easy to use on computers without graphics support or NON-GPU computers. In this research, a vehicle counter system is built by implementing pre-trained object detection model, MobileNet SSD and object tracking algorithm MOSSE (Minimum Output Sum of Squared Error). The model and algorithm are implemented with a library for processing digital images, namely OpenCV to build a vehicle counter system with a high comparison value of the predicted number of vehicles and the actual value of the number of vehicles, has low computing power, and can be run on a NON-GPU computer. Keywords: Vehicle Counter System, OpenCV, SSD, MobileNet, MOSSE.
UNJUK KERJA PENGGUNAAN CONVOLUTIONAL NEURAL NETWORK (CNN) DAN SIAMESE NEURAL NETWORK (SNN) DALAM MENGIDENTIFIKASI WAJAH Airlangga, Daniar Putri; Mudjirahardjo, Panca; Rahmadwati, n/a
Jurnal Mahasiswa TEUB Vol. 12 No. 1 (2024)
Publisher : Jurnal Mahasiswa TEUB

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Face identification is one form of security verification that is currently widely used due to its low level of fraud. Therefore, this study aims to obtain the best model with the method and parameters between CNN and SNN in identifying faces using a dataset sourced from the open-source Kaggle created by Burak from Middle East Technical University, Turkey. The dataset consists of 5 classes, each containing 5 faces with different names but the same number of faces per class. Furthermore, the existing data will undergo a preprocessing process aimed at reducing dimensions and noise in non-uniform data. Next, in this process, the data will be divided into training data and test data for further training using the Tensorflow Keras framework and the visual studio code IDE and Kaggle as a notebook. There are 2 stages of testing in this study, namely testing the performance of the modeling on the system created by using varying epochs and changes in parameters, and testing the performance of the system in performing identification. The model with the best results is the SNN method model, which has a training accuracy of up to 99% and an evaluation result accuracy of up to 98% with a prediction time of only 0.7 seconds. The results of this study indicate that the SNN method has better ability in identifying faces compared to the CNN method. Keywords— Face identification, Machine Learning, Convolutional Neural Network (CNN), Siamese Neural Network (SNN).
DETEKSI BANGUNAN MENGGUNAKAN CONVOLUTIONAL NEURAL NETWORK (CNN) ARSITEKTUR U-NET PADA CITRA SATELIT Firmansyah, Vicky; Mudjirahardjo, Panca; Rahmadwati, n/a
Jurnal Mahasiswa TEUB Vol. 12 No. 3 (2024)
Publisher : Jurnal Mahasiswa TEUB

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

This research aims to detect buildings in satellite images using the U-Net architecture CNN model. The dataset is a satellite image of the Araya area, Malang City extracted through Google Earth Engine with the Python programming language, consisting of 4 band channels: Red, Green, Blue, and NIR. The data is processed into 2 types of sizes, namely 128x128 pixels and 64x64 pixels, then divided into train, validation, and test data. Training uses Keras Tensorflow, and testing is done in three stages: model performance testing with layer addition and parameter changes, building detection testing, and comparison of prediction results with manual measurements. The best model was obtained by a model with a training dataset of 128x128 pixels that achieved 93%accuracy with a prediction time of 2 seconds. Keywords— Building Detection, Semantic Segmentation, Convolutional Neural Network (CNN), UNet, Satellite Imagery
IDENTIFIKASI JENIS TANAH MENGGUNAKAN CONVOLUTIONAL NEURAL NETWORK (CNN) DENGAN ARSITEKTUR RESIDUAL NETWORK (RESNET-50) DAN MOBILE NETWORK (MOBILENETV2) Syarifah, Naily; Mudjirahardjo, Panca; Razak, Angger Abdul
Jurnal Mahasiswa TEUB Vol. 12 No. 3 (2024)
Publisher : Jurnal Mahasiswa TEUB

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

The research was conducted to identify soil types using artificial intelligence using the Convolutional Neural Network method.(CNN). This is done to help young farm activists stay up-to-date to get land use information, i.e. by helping in optimizing theidentification of land or land as a growing medium. The study uses the MobileNetV2 and ResNet-50 architectures to identify different soil types. Both architectures compared their performance in identifying soil types through the texture and colour taken on the test image set data. Before doing the training of course the data will be used through pre-processing for the consistency of input and maximize the modeling process. Both were tested by performing several scenarios to obtain each the best performance results of the optimizer, the number of epochs and the learning rate values. Models of both architectures have a high degree of accuracy and precision. 3. For the MobileNet architecture, the V2 produced models with accuracy values of 91.91%, loss of 90.87%, and a prediction time of 0.108 seconds. And for the ResNet-50 architecture the model produced a precision value of 99.08%, precision 99.11%, recall 99.12%, F1-score 99.10%, specification 99.85%, loss 5.02% and forecast time of 0.037 seconds. Keywords: CNN, Soil Classification, MobileNetV2, ResNet-50
PERANCANGAN PERANGKAT LUNAK BASE STATION ROBOT SEPAK BOLA BERODA DENGAN QT CREATOR Nathanael, Indra; Mudjirahardjo, Panca; Muttaqin, Adharul
Jurnal Mahasiswa TEUB Vol. 12 No. 5 (2024)
Publisher : Jurnal Mahasiswa TEUB

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

This research aims to design base station software used in the Indonesian Wheeled Soccer Robot Contest (KRSBI-B) competition. The base station plays a role in the KRSBI-B communication scheme, which is to receive commands from the referee box and send them to the client. In this study, the base station is designed using Qt Creator as an IDE on the Ubuntu Operating System (OS). In general, the base station is divided into 3 main points, namely communication with the referee box, communication with the robot/client, and an interface to interact with the user. Communication between the base station and the referee box is designed using the TCP/IP protocol. Communication between the base station and the client is designed using the UDP protocol. The basestation interface is designed using the Qt Framework as the backend and Qwidget as the output. In this study, testing was also carried out consisting of functionality testing and performance testing. Functionality testing is done to test whether the base station can work, while performance testing is done to test the performance of the base station during use. Based on the test results, the designed base station has successfully received data from the referee box and sent and received data to the clientcorrectly. The designed base station is ready to use and meets the requirements for the KRSBI-B competition. Keywords— Base Station, Qt Creator, TCP/IP, UDP.
Partial half fine-tuning for object detection with unmanned aerial vehicles Pebrianto, Wahyu; Mudjirahardjo, Panca; Pramono, Sholeh Hadi
IAES International Journal of Artificial Intelligence (IJ-AI) Vol 13, No 1: March 2024
Publisher : Institute of Advanced Engineering and Science

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.11591/ijai.v13.i1.pp399-407

Abstract

Deep learning has shown outstanding performance in object detection tasks with unmanned aerial vehicles (UAVs), which involve the fine-tuning technique to improve performance by transferring features from pre-trained models to specific tasks. However, despite the immense popularity of fine-tuning, no works focused on to study of the precise fine-tuning effects of object detection tasks with UAVs. In this research, we conduct an experimental analysis of each existing fine-tuning strategy to answer which is the best procedure for transferring features with fine-tuning techniques. We also proposed a partial half fine-tuning strategy which we divided into two techniques: first half fine-tuning (First half F-T) and final half fine-tuning (Final half F-T). We use the VisDrone dataset for the training and validation process. Here we show that the partial half fine-tuning: Final half F-T can outperform other fine-tuning techniques and are also better than one of the state-of-the-art methods by a difference of 19.7% from the best results of previous studies.
Pendeteksi Gawang dan Bola Hasil Tangkapan Kamera Omnidirectional Menggunakan YOLOv4-TINY Rauf, Daru Adiyatma; Mudjirahardjo, Panca; Rahmadwati, n/a
Jurnal Mahasiswa TEUB Vol. 13 No. 2 (2025)
Publisher : Jurnal Mahasiswa TEUB

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Penelitian ini bertujuan untuk merancang sistem pendeteksi gawang dan bola pada robot sepak bola beroda menggunakan kamera omni-directional sebagai sensor visual utama. Sistem ini ditujukan untuk mendukung performa robot dalam ajang Kontes Robot Indonesia (KRI) divisi Kontes Robot Sepak Bola Indonesia Beroda (KRSBI-B). Kamera omni-directional digunakan untuk memperoleh citra dengan sudut pandang luas sehingga dapat memantau lingkungan sekitar secara menyeluruh dalam satu frame. Metode You Only Look Once (YOLOv4-Tiny) dipilih karena kemampuannya dalam mendeteksi objek secara real-time dengan kecepatan dan akurasi yang baik, sesuai dengan kebutuhan kompetisi robotika. Tahapan penelitian meliputi akuisisi data citra menggunakan kamera omni-directional, pre-processing citra untuk meningkatkan kualitas gambar, serta pelatihan model YOLOv4-Tiny untuk mendeteksi objek gawang dan bola. Selanjutnya, sistem diintegrasikan ke dalam robot menggunakan bahasa pemrograman C++ dengan bantuan pustaka OpenCV dan Darknet untuk memproses citra secara efisien. Pada penelitian ini, dilakukan juga pengujian yang terdiri dari pengujian evaluasi model, pengujian waktu komputasi , dan pengujian deteksi objek di luar area lapangan. Pengujian evaluasi model dilakukan untuk mengevaluasi kemampuan model dalam mendeteksi objek secara tepat menggunakan metrik evaluasi. Sementara itu, pengujian performansi untuk menghitung waktu inferensi untuk satu frame. Berdasarkan hasil pengujian, Model YOLOv4-Tiny mampu mencapai AP pada kelas Bola sebesar 94.16%, AP pada kelas Gawang sebesar 100%, dengan mAP sebesar 97.08%. Selain itu, model ini memiliki Recall sebesar 0.97, Precision sebesar 0.94, dan F1-Score sebesar 0.96. Dari segi waktu komputasi, sistem memiliki waktu inferensi untuk satu frame di antara 10 ms hingga 15 ms. Hasil pengujian di luar area lapangan menunjukkan bahwa akurasi deteksi turut dipengaruhi oleh kondisi lingkungan seperti warna latar belakang, pencahayaan, dan perspektif kamera. Objek dengan kontras tinggi lebih mudah terdeteksi, sedangkan objek yang menyatu dengan latar sulit dikenali. Jarak dan distorsi kamera omni-directional juga memengaruhi performa deteksi. Secara keseluruhan, model memiliki precision 0,86, recall 0,95, dan F1-score 0,90, yang tetap menunjukkan performa deteksi yang cukup seimbang di kondisi yang lebih menantang. Dengan hasil tersebut, model ini mampu mendeteksi gawang dan bola dengan tingkat akurasi yang memuaskan pada berbagai kondisi posisi objek. Sistem ini memiliki waktu pemrosesan yang cukup cepat sehingga mendukung performa robot dalam pengambilan keputusan secara real-time. Sistem deteksi gawang dan bola sudah layak digunakan dan memenuhi syarat pada pertandingan KRSBI-B. Kata Kunci— Kamera Omni-Directional, YOLOv4-Tiny, Deteksi Gawang, Deteksi Bola
Tongkat Multi-Sensor untuk Orang dengan Gangguan Penglihatan Hartono, Rafendra Ariwardana; Muslim, Muhammad Aziz; Mudjirahardjo, Panca
Jurnal Mahasiswa TEUB Vol. 13 No. 3 (2025)
Publisher : Jurnal Mahasiswa TEUB

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Tongkat putih konvensional merupakan tongkat umum yang digunakan  oleh orang dengan gangguan penglihatan/tunanetra untuk mendeteksi  rintangan disekitarnya. Namun, rintangan yang dideteksi hanya berupa  rintangan yang berada di tanah atau yang bersentuhan dengan ujung  tongkat. Oleh karena itu, penelitian ini bertujuan untuk merancang dan mengembangkan tongkat multi-sensor yang bisa mendeteksi rintangan di  atas, depan, bawah, serta deteksi keberadaan air berbasis mikrokontroller  Arduino yang dilengkapi dengan sensor ultrasonik, sensor LIDAR, sensor  deteksi air, buzzer dan getaran motor. Sistem ini dirancang untuk memberi peringatan berupa suara maupun getaran apabila sensor mendeteksi  adanya rintangan atau genangan air. Metode penelitian yang digunakan  meliputi studi literatur, perancangan sistem, implementasi perangkat, dan  pengujian performa sensor di berbagai skenario navigasi, baik di  lingkungan indoor maupun outdoor. Hasil pengujian menunjukkan bahwa  sistem mampu mendeteksi rintangan dengan tingkat akurasi tinggi,  dengan error pengukuran sensor ultrasonik sekitar 0,27% dan sensor  LIDAR rata-rata error sebesar 0,23% hingga 1,25% tergantung jarak pengujian. Selain itu, evaluasi persepsi pengguna menunjukkan bahwa tongkat multi-sensor ini mampu meningkatkan rasa aman, kenyamanan,  dan mobilitas penyandang tunanetra dalam beraktivitas sehari-hari.  Dengan biaya pembuatan yang relatif rendah, alat ini diharapkan dapat  menjadi solusi inovatif yang terjangkau dan aplikatif untuk meningkatkan kualitas hidup penyandang tunanetra. Kata Kunci: Tongkat multi-sensor, tunanetra, Arduino Uno, sensor  ultrasonik, sensor LIDAR, sensor deteksi air.
IMPLEMENTASI YOLO DALAM SISTEM PENGAWASAN UJIAN DENGAN PROSES DETEKSI SECARA REAL-TIME Brahmana, Nigel Shidqy Razendriya; Mudjirahardjo, Panca; Razak, Angger Abdul
Jurnal Mahasiswa TEUB Vol. 13 No. 3 (2025)
Publisher : Jurnal Mahasiswa TEUB

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Tingkat kecurangan dalam pelaksanaan ujian di Indonesia masih menjadi permasalahan serius, yang semakinkompleks akibat kemajuan teknologi dan kemudahan akses perangkat elektronik seperti ponsel dan laptop. Berdasarkanlaporan Kemendikbud (2019), terdapat 202 aduan kecurangan dalam Ujian Nasional dengan 126 kasus terkonfirmasi, sertasurvei oleh Winardi & Anggraeni (2017) menunjukkan bahwa 77,5% mahasiswa akuntansi mengakui pernah melakukankecurangan akademik. Penelitian ini mengimplementasikan algoritma YOLOv5 dalam sistem pengawasan ujian untukmendeteksi objek terlarang secara real-time. Sistem ini dirancang untuk mengidentifikasi keberadaan barang-barang sepertiponsel, tas, dan buku yang dapat mengindikasikan kecurangan selama ujian. YOLOv5 dipilih karena efisiensi dan akurasinyadalam deteksi objek. Metode penelitian melibatkan pengumpulan data citra, pelabelan objek, pelatihan model YOLOv5, danpengujian kinerja sistem. Hasil pengujian menunjukkan bahwa sistem mampu mendeteksi objek terlarang dengan baik,dengan nilai precision 0,84, recall 0,791, dan mAP 0,852. Sistem ini berpotensi untuk diintegrasikan ke dalam kegiatan ujianuntuk meningkatkan pengawasan ujian secara otomatis.Kata Kunci— YOLOv5, Sistem Proctoring, Deteksi Objek, Computer Vision, Real-time.
IDENTIFIKASI NON-INVASIF SKOLIOSIS MENGGUNAKAN METODE CONVOLUTIONAL NEURAL NETWORK (CNN) DENGAN ARSITEKTUR RESIDUAL NETWORK (RESNET18) DAN MOBILE NETWORK (MOBILENETV2) Izanati, Nazuha; Mudjirahardjo, Panca; Setyawan, Raden Arief
Jurnal Mahasiswa TEUB Vol. 13 No. 3 (2025)
Publisher : Jurnal Mahasiswa TEUB

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Skoliosis merupakan kelainan tulang belakang yang menyebabkan kelengkungan abnormal dan berdampak padapostur serta kesehatan jangka panjang penderitanya. Metode deteksi konvensional seperti X-ray memiliki risiko paparan radiasiserta keterbatasan akses di beberapa wilayah. Penelitian ini bertujuan untuk mengidentifikasi skoliosis secara non-invasifmelalui citra tampak belakang tubuh manusia menggunakan model Convolutional Neural Network (CNN) dengan duapendekatan arsitektur, yaitu ResNet-18 dan MobileNetV2. Dataset citra yang digunakan terdiri dari dua kelas: normal danskoliosis, dan diolah melalui beberapa skenario eksperimental seperti representasi warna, jumlah epoch, jenis optimizer, sertanilai learning rate. Hasil menunjukkan bahwa MobileNetV2 memberikan performa terbaik dengan akurasi 88% dan waktukomputasi yang lebih efisien dibandingkan ResNet-18 yang mencapai akurasi 84%. Perbandingan menunjukkan bahwaMobileNetV2 unggul dalam efisiensi, sementara ResNet-18 lebih konsisten dalam performa klasifikasi. Pendekatan inimembuktikan bahwa CNN dapat menjadi solusi alternatif yang aman, efisien, dan mudah diterapkan untuk deteksi skoliosissecara dini dan non-invasif.Kata Kunci—Skoliosis, CNN, ResNet-18, MobileNetV2, Klasifikasi Citra, Deteksi Non-Invasif
Co-Authors Abdul Goffar Ricky Mahendra Achmad Basuki Adharul Muttaqin Ahmad Syafiq Kanzul Fikri Aiman Muhamad Basymeleh Airlangga, Daniar Putri Alkafi Dimitri Sukmana Andy Kurnia Santoso Angger Abdul Razak Anthony Wijoyo Arafah, Ghifari Raihan Bagus Esa Pramudya Bidin Yuniar Hamzah Bima Feridhan Nugraha Bimasena, Muhammad Farrel Brahmana, Nigel Shidqy Razendriya Chandra Halim Harahap Dachlan, Hary Soekotjo Danny Kurnianto Doni Juli Wiranata Eka Maulana Erni Yudaningtyas Esa Ilham Akbar Faradisa , Annisa' Illah Farihah Hedar Fatchur Rozi Al Fitrah Fauzi, Maher Feishal Reza Firmansyah, Vicky Gilang Luih Pinandita Haidar Taqy Hartono, Rafendra Ariwardana Hary Soekotjo Dachlan Hasdi Sasandi Ismail Musirin Ismail Musirin Ita Dwi Purnamasari Izanati, Nazuha Juan Mora Michael Marbun Juli Arianes Leonard Dimas Prakoso Lilik J. Awalin Lukman Gumelar M Fauzan Edy Purnomo M. Hanif Azhary M. Julius St Julius St M. Julius St Julius St, M. Julius St Machfud Firmansyah Manerep Luis Fernando Purba Marco Gunawan Maulana, Eka Miladina Rizka Aziza Mohammad Alif Robby Gani Mohammad Ilhammudin Toiyib Monifa Arini Muhammad Akbar Muhammad Aziz Muslim Muhammad Ikhsan Muhammad Ivan Fadillah Muhammad Rafi’ Zaidan Maajid n/a Soeprapto Nanang Sulistiyanto Nathanael, Indra Notario Pramudita Nugraha, Dimas Aji Nurus Sa'adah Octarudin Mahendra Oky Risky Dwi Santoso Pangemanan, Christofel Panjaitan, Gian Amadea Pebrianto, Wahyu Permatasari, Alissa Dyah Ayu Ponco Siwindarto Pratolo Rahardjo R. A. Setyawan Rachmawati, Luthfiyah Raden Arief Setyawan Rahmadwati Rahmadwati Rahmadwati, n/a Rahmadwati, n/a Rahmadwati, Rahmadwati Rauf, Daru Adiyatma Reinato Teguh Santoso Reza, Feishal Ricky Insyani Santosa P. P. Ridho Herasmara Rif'an, Mochammad Rifqa Asruroh Efnif Rini Nur Hasanah Riza Hasbi Ash Shiddieqy Rizky Aiman Haniffalah Harijanto Robbith Qosath Al Auhi Rohman, Muhammad Ariefur Samuel Aji Sena Sena, Samuel Aji Septi Uliyani Sholeh Hadi Pramono Sirojul Hadi Sofyan Andika Yusuf Sultoni Sultoni Sultoni, Sultoni Surya Agung Kurnia Suyono, Hadi Syarifah, Naily Tri Nurwati Vira Zafarin Waru Djuriatno Waru Djuriatno Wuri Roro Indraswari yuliana diah pristanti Zainuri, Akhmad