Hasil Pencarian  ::  Simpan CSV :: Kembali

"

Data LiDAR banyak menggantikan data dua dimensi untuk merepresentasikan data geografis karena kekayaan informasi yang dimilikinya. Salah satu jenis pemrosesan data LiDAR adalah segmentasi semantik tutupan lahan yang mana telah banyak dikembangkan menggunakan pendekatan model deep learning. Algoritma-algoritma tersebut menggunakan representasi jarak Euclidean untuk menyatakan jarak antar poin atau node. Namun, sifat acak dari data LiDAR kurang sesuai jika representasi jarak Euclidean tersebut diterapkan. Untuk mengatasi ketidaksesuaian tersebut, penelitian ini menerapkan representasi jarak non-Euclidean yang secara adaptif diupdate menggunakan nilai kovarian dari set data point cloud. Ide penelitian ini diaplikasikan pada algoritma Dynamic Graph Convolutional Neural Network (DGCNN). Dataset yang digunakan dalam penelitian ini adalah data LiDAR Kupang. Metode pada penelitian ini menghasilkan performa nilai akurasi 75,55%, di mana nilai akurasi ini lebih baik dari algoritma dasar PointNet dengan 65,08% dan DGCNN asli 72,56%. Peningkatan performa yang disebabkan oleh faktor perkalian dengan invers kovarian dari data point cloud dapat meningkatkan kemiripan suatu poin terhadap kelasnya.

---

LiDAR data widely replaces two-dimensional geographic data representation due to its information resources. One of LiDAR data processing tasks is land cover semantic segmentation which has been developed by deep learning model approaches. These algorithms utilize Euclidean distance representation to express the distance between the points. However, LiDAR data with random properties are not suitable to use this distance representation. To overcome this discprepancy, this study implements a non-Euclidean distance representation which is adaptively updated by applying their covariance values. This research methodology was then implemented in Dynamic Graph Convolutional Neural Network (DGCNN) algorithm. The dataset in this research is Kupang LiDAR. The results obtained performance accuracy value of 75.55%, which is better than the baseline PointNet of 65.08% and Dynamic Graph CNN of 72.56%. This performance improvement is caused by a multiplication of the inverse covariance value of point cloud data, which raised the points similarity to the class.

"

"Skripsi ini membahas mengenai penggunaan model segmentasi semantik UNet sebagai alternatif metode segmentasi wajah dan tangan gerakan isyarat SIBI (Sistem Isyarat Bahasa Indonesia) pada latar belakang kompleks. Penelitian dilakukan terhadap dataset gerakan isyarat SIBI milik Lab MLCV Fakultas Ilmu Komputer Universitas Indonesia. Dalam penelitian ini, dilakukan percobaan dengan tiga jenis konfigurasi UNet, yaitu UNet 4- level tanpa Batch Normalization, UNet 5-level tanpa Batch Normalization, dan UNet 4- level dengan Batch Normalization. Hasil segmentasi dari UNet konfigurasi terbaik kemudian dilakukan tahap pengenalan selanjutnya, yaitu ekstraksi fitur dengan MobileNetV2, penghapusan gerakan transisi dengan TCRF, dan gesture recognition dengan 2-layer biLSTM untuk mendapatkan hasil translasi serta evaluasi akhir. Selain itu, performa sistem dengan menggunakan metode segmentasi UNet dibandingkan dengan performa sistem dengan menggunakan metode segmentasi RetinaNet+Skin Color Segmentation. Hasil dari penelitian didapatkan bahwa konfigurasi UNet 4-level dengan Batch Normalization menghasilkan segmentasi yang sedikit lebih baik dibandingkan konfigurasi lainnya, yaitu dengan nilai IOU 0,9178% pada dataset berlatar belakang kompleks. Performa UNet terlihat baik pada saat kedua tangan berada di depan badan, dan menurun ketika tangan berada di posisi yang berdekatan dengan area kulit lainnya (lengan, leher, wajah). Didapatkan juga bahwa sistem pengenalan isyarat SIBI ke teks bahasa Indonesia dengan menggunakan metode segmentasi UNet berhasil memiliki performa yang lebih baik dibandingkan menggunakan metode segmentasi RetinaNet+Skin Color Segmentation, dengan nilai WER 2,703% dan SAcc 82,424% pada latar belakang kompleks. Didapatkan juga waktu komputasi UNet yang lebih cepat dibandingkan RetinaNet dengan waktu segmentasi 0,19643 detik per frame pada CPU NVIDIA DGX A100

---

This thesis discusses the use of the UNet semantic segmentation model as an alternative to hand and face segmentation methods for SIBI (Indonesian Signing System) on complex backgrounds. This research was conducted on SIBI gesture dataset by MLCV Lab (Faculty of Computer Science, Universitas Indonesia). In this study, experiments were conducted with three types of UNet configurations, namely 4-level UNet without Batch Normalization, 5-level UNet without Batch Normalization, and 4-level UNet with Batch Normalization. Segmentation results from the best UNet configuration is then carried out in the next stage of the system, namely feature extraction with MobileNetV2, epenthesis removal with TCRF, and gesture recognition with 2-layer biLSTM to obtain translation results and the final evaluations. In addition, system performance using the UNet segmentation method is compared to system performance using the RetinaNet+Skin Color Segmentation method. The results of the study showed that the 4-level UNet configuration with Batch Normalization produces slightly better segmentation than the other configurations, with an IOU of 0.9178% on a dataset with a complex background. Based on the sample results, UNet performance is good when both hands are on the front of the body, and it decreases when the hands are in close proximity to other skin areas (arms, neck, face). It was also found that the SIBI gesture recognition system to Indonesian text using the UNet segmentation method managed to have better performance than using the RetinaNet+Skin Color Segmentation, with a WER value of 2.703% and a SAcc of 82.424% on a complex background. It was also found that UNet processing time was faster than RetinaNet with a segmentation rate of 0.19643 seconds per frame on the NVIDIA DGX A100 CPU."

"Kuantifikasi standar lemak jantung menggunakan citra nonkontras dapat menjadi suatu nilai prognostik tambahan dalam mengevaluasi penyakit jantung koroner. Metode otomatis berbasis deep learning memiliki kelebihan dari metode manual yaitu mengurangi waktu kuantifikasi, beban kerja dan user dependence. Pada penelitian ini, lemak jantung epikardial dan mediastinal dari dataset open source dan dari Rumah Sakit Mayapada Tangerang disegmentasi menggunakan segmentasi semantik berbasis CNN DeepV3+ Resnet18 dan dievaluasi. Volume dari lemak jantung diestimasikan menggunakan fitur regionprops Matlab 2021a. Sistem dapat segmentasi lemak jantung pada keakurasian tertinggi sebesar 98,8 % dan dice score sebesar 0,76 untuk lemak epikardial dan keakurasian 96,8% dan dice score sebesar 0,69 untuk lemak mediastinal dataset open source. Namun, pada data uji yaitu data CT jantung yang diambil dari rumah sakit menghasilan keakurasian tertinggi pada 28% untuk lemak epikardial. Secara kualitatif, struktur seperti lemak abdomen, otot jantung dan tulang belakang masih ikut tersegmen. Setelah melakukan penyesuaian citra antara data uji dengan data pelatihan, akurasi tertinggi pada lemak epikardial sebesar 97%. Namun, lemak epikardial dan mediastinal belum berhasil untuk dipisahkan. Volume lemak jantung untuk kedua dataset berhasil diestimasikan. Metode volume manual dengan metode otomatis menunjukkan korelasi yang kuat (R2= 0,9843) dengan standard error sebesar 3,86 namun terlihat bahwa terjadi eror sistematik.

---

Standard quantification of cardiac fat using non-contrast images can be additional prognostic value in evaluating coronary heart disease. Automatic methods based on deep learning have advantages over manual methods, namely reducing quantification time, workload and user dependence. In this study, epicardial and mediastinal cardiac fat from open source dataset and Mayapada Hospital Tangerang were segmented using CNN DeepV3+ Resnet18-based semantic segmentation and evaluated. The volume of cardiac fat was estimated using the regionprops feature of Matlab 2021a. The system can segment cardiac fat at the highest accuracy of 98.8% and a dice score of 0.76 for epicardial fat and 96.8% accuracy and a dice score of 0.69 for mediastinal fat of the open source dataset. However, the test dataset, namely cardiac CT data taken from the hospital, yielded the highest accuracy at 28% for epicardial fat. Qualitatively, structures such as abdominal fat, cardiac muscle and spine are still segmented. After adjusting the image between the test data and the training data, the highest accuracy in epicardial fat was 97%. However, epicardial and mediastinal fat have not been successfully separated. Heart fat volumes for both datasets were successfully estimated. The manual volume method in respect to the automatic method showed a strong correlation (R2= 0.9843) with a standard error of 3.86, but it was seen that there was a systematic error."

"This book reports on advanced theories and methods in three related fields of research: applied physics, system science and computers. It is organized in three parts, the first of which covers applied physics topics, including lasers and accelerators; condensed matter, soft matter and materials science; nanoscience and quantum engineering; atomic, molecular, optical and plasma physics; as well as nuclear and high-energy particle physics. It also addresses astrophysics, gravitation, earth and environmental science, as well as medical and biological physics. The second and third parts focus on advances in computers and system science, respectively, and report on automatic circuit control, power systems, computer communication, fluid mechanics, simulation and modeling, software engineering, data structures and applications of artificial intelligence among other areas. Offering a collection of contributions presented at the 2nd International Conference on Applied Physics, System Science and Computers (APSAC), held in Dubrovnik, Croatia on September 27–29, 2017, the book bridges the gap between applied physics and electrical engineering. It not only to presents new methods, but also promotes collaborations between different communities working on related topics at the interface between physics and engineering, with a special focus on communication, data modeling and visualization, quantum information, applied mechanics as well as bio and geophysics."