Jurnal Ilmu Komputer dan Informasi
Not a member yet
247 research outputs found
Sort by
Improved mask RCNN and cosine similarity using RGBD segmentation for Occlusion handling in Multi Object Tracking
Full text linkIn this study, additional depth images were used to enrich the information in each image pixel. Segmentation, by its nature capable to process image up to pixel level. So, it can detect up to the smallest part of the object, even when it’s overlapped with another object. By using segmentation, the main goal is to be able to maintain the tracking process longer when the object starts to be occluded until it is severely occluded right before it is completely disappeared. Object tracking based on object detection was developed by modifying the Mask R-CNN architecture to process RGBD images. The detection results feature extracted using HOG, and each of them got compared to the target objects. The comparison was using cosine similarity calculation, and the maximum value of the detected object would update the target object for the next frame. The evaluation of the model was using mAP calculation. Mask R-CNN RGBD late fusion had a higher value by 5% than Mask R-CNN RGB. It was 68,234% and 63,668%, respectively. Meanwhile, the tracking evaluation uses the traditional method of calculating the id switching during the tracking process. Out of 295 frames, the original Mask R-CNN method had ten switching ID times. On the other hand, the proposed method Mask R-CNN RGBD had much better tracking results with switching ids close to 0. Keywords—Occlusion, RGBD, Mask R-CNN, Late fusion, Cosine similarit
Face Spoofing Detection using Inception-v3 on RGB Modal and Depth Modal
Full text linkFace spoofing can provide inaccurate face verification results in the face recognition system. Deep learning has been widely used to solve face spoofing problems. In face spoofing detection, it is unnecessary to use the entire network layer to represent the difference between real and spoof features. This study detects face spoofing by cutting the Inception-v3 network and utilizing RGB modal, depth, and fusion approaches. The results showed that face spoofing detection has a good performance on the RGB and fusion models. Both models have better performance than the depth model because RGB modal can represent the difference between real and spoof features, and RGB modal dominate the fusion model. The RGB model has accuracy, precision, recall, F1-score, and AUC values obtained respectively 98.78%, 99.22%, 99.31.2%, 99.27%, and 0.9997 while the fusion model is 98.5%, 99.31%, 98.88%. 99.09%, and 0.9995, respectively. Our proposed method with cutting the Inception-v3 network to mixed6 successfully outperforms the previous study with accuracy up to 100% using the MSU MFSD benchmark dataset
A Hybrid Virtual Assistant for Legal Domain Based on Information Retrieval and Knowledge Graphs
Full text linkVirtual assistants have gained popularity across various domains, including the legal field, where they serve to offer guidance and aid in the form of law retrieval. In this research, our aim is to develop a legal virtual assistant that combines knowledge graphs (KGs) and information retrieval (IR) techniques. This hybrid approach allows us to provide accurate answers extracted from structured interconnected data while simultaneously cater to a diverse range of legal inquiries. We categorize these inquiries into a few distinct use cases: definition lookup, law component lookup, sanctions, and domain knowledge. Our system encompasses a chatbot platform, knowledge graph querying, and information retrieval. Specifically, we construct a VA system over a legal knowledge graph pertaining to the Indonesian Act concerning Manpower or Labor (UU Ketenagakerjaan) and the Indonesian Act concerning the Creation of Jobs (UU Cipta Kerja). This marks the creation of the first legal virtual assistant in the Indonesian context that combines KG and IR methodologies. To evaluate the effectiveness of our prototype system, we conduct tests using a variety of labor law-related questions, ranging in difficulty. The integration of knowledge graphs and information retrieval proves to significantly improve the support provided for a wide range of potential applications in the legal field
Encoder-Decoder with Atrous Spatial Pyramid Pooling for Left Ventricle Segmentation in Echocardiography
Full text linkAssessment of cardiac function using echocardiography is an essential and widely used method. Assessment by manually labeling the left ventricle area can generally be time-consuming, error-prone, and has interobserver variability. Thus, automatic delineation of the left ventricle area is necessary so that the assessment can be carried out effectively and efficiently. In this study, encoder-decoder based deep learning model for left ventricle segmentation in echocardiography was developed using the effective CNN U-Net encoder and combined with the deeplabv3+ decoder which has efficient performance and is able to produce sharper and more accurate segmentation results. Furthermore, the Atrous Spatial Pyramid Pooling module were added to the encoder to improve feature extraction. Tested on the Echonet-Dynamic dataset, the proposed model gives better results than the U-Net, DeeplabV3+, and DeeplabV3 models by producing a dice similarity coefficient of 92.87%. The experimental results show that combining the U-Net encoder and DeeplabV3+ decoder is able to provide increased performance compared to previous studies
Poetry Generation for Indonesian Pantun: Comparison Between SeqGAN and GPT-2
Full text linkPantun is a traditional Malay poem consisting of four lines: two lines of deliverance and two lines of messages. Each ending-line word in pantun forms an ABAB rhyme pattern. In this work, we compare the performance of Sequence Generative Adversarial Nets (SeqGAN) and Generative Pre-trained Transformer 2 (GPT-2) in automatically generating Indonesian pantun. We also created the first publicly available Indonesian pantun dataset that consists of 7.8K pantun. We evaluated how well each model produced pantun by its lexical richness and its formedness. We introduced the evaluation of pantun with two aspects: structure and rhyme. GPT-2 performs better with a margin of 29.40% than SeqGAN in forming the structure, 35.20% better in making rhyming patterns, and 0.04 difference in giving richer vocabulary to its generated pantun
LexID: The Metadata and Semantic Knowledge Graph Construction of Indonesian Legal Document
Full text linkThe Legal Fiction principle stipulates that the government needs to ensure the public availability of all of their legal documents. Unfortunately, the text-based search services they provide cannot return satisfactory answers in retrieval scenarios requiring proper representation of relationships between various legal documents. A key problem here is the lack of explicit representation of such relationships behind the employed retrieval engines. We aim to address this problem by proposing LexID knowledge graph (KG) that provides an explicit knowledge representation for Indonesian legal domain usable for such retrieval purposes. The KG contains both legal metadata information and semantic content of the legal clauses of the legal document's articles, modeled using formal vocabulary from the LexID ontology also presented in this paper. The KG is constructed from thousands of Indonesian legal documents. Since the procedure of writing a legal document regulated by the government is clear and detailed, we use a rule-based approach to construct our KG. At the end, we describe several use cases of the KG to address different retrieval needs. In Addition, we evaluated the quality of our KG by measuring its ability to answer questions and got that LexID can answer questions with the macro average F1 score is about 0.91
Knowledge Management for Electronic-Based Government System Using Semantic Thesaurus
Full text linkSistem Pemerintahan Berbasis Elektronik (SPBE), the electronic-based government system, is Indonesia’s egovernment policy providing services to citizens through information and communication technology. Knowledge containing SPBE must be managed in various ways, one of which is the creation of an SPBE thesaurus to facilitate access and search for SPBE-related items using words or terms about it. In this study, we provide an overview of the thesaurus development process that has complied with the ISO 25964 standard and uses the Simple Knowledge Organization System (SKOS) as the application of the thesaurus in the web environment. Basic concepts or related terms and relationships between concepts have been linked with similar concepts in other thesauri that have existed before. This research also looks at the process of automating the recognition of related terms in internet articles using Word2Vec and Doc2Vec. In the process of adding terms, we discover challenges in filtering terms, determining relationships between terms, and determining reciprocal relationships between terms
Fine Tuning of Interval Configuration for Deep Reinforcement Learning Based Congestion Control
Full text linkIt is apparent that various internet services in today’s digital ecosystem effectuate different types of networks’ quality of services (QoS) requirements. This condition, in fact, adds another level of complexity to the current network congestion control protocols. Therefore, it drives the adoption of deep reinforcement learning to improve the protocols’ adaptability to the dynamic networks’ QoS requirements. In this case, the state-of-the-art works on congestion control protocols, formulate the markov decision process (MDP) by transforming the congestion control pattern from the saw tooth congestion window to the staircase sending rate per-interval cycles. This approach treats congestion control as a sequential decision-making process that fits reinforcement learning. However, the interval configuration parameter that gives the optimum QoS has not been empirically studied. In this work, we present an extensive study on various interval configuration parameters for the deep reinforcement learning-based congestion control agent. Our work shows that various interval configuration, which consists of the RTT estimator and the n parameter, results in different QoS. The experiment shows that the RTTjk has significantly higher throughput than RTTewma and RTTmin−filtered in various network conditions. Furthermore, we found that the RTTjk with n = 2.0 is superior to other configurations in almost all networking scenarios. Whereas the RTTjk with n = 1.0 is optimal for a network environment with fixed bandwidth scenario
Yoga Pose Rating using Pose Estimation and Cosine Similarity
Full text linkOne type of exercise that many people do today is yoga. However, doing yoga yourself without an instructor carries a risk of injury if not done correctly. This research proposes an application in the form of a website that can assess the accuracy of a person's yoga position, by using ResNet for pose estimation and cosine similarity for calculating the similarity of positions. The application will recognize a person's body pose and then compare it with the poses of professionals so that the accuracy of their position can be assessed. There are three types of datasets used, the first is the COCO dataset to train a pose estimation model so that it can recognize someone's pose, the second is a reference dataset that contains yoga poses performed by professionals, and the third is a dataset that contains pictures of yoga poses that are considered correct. There are 9 yoga poses used, namely Child's Pose, Swimmers, Downdog, Chair Pose, Crescent Lunge, Planks, Side Plank, Low Cobra, Namaste. The optimal pose estimation model has a precision value of 87% and a recall of 88.2%. The model was obtained using the Adam optimizer, 30 epochs, and a learning rate of 0.0001
Biometric System for Person Authentication Using Retinal Vascular Branching Pattern
Full text linkThe person’s retina has its uniqueness that can be used as biometric recognition. The use of the retina as a marking feature in biometrics is more accurate in making calls, verification, and authentication. Retinal biometric characteristics are unique and difficult to manipulate, thus making the retinal biometric system one of the most reliable biometrics compared to other biometric characteristics. The retinal biometric system can be formed using extracted retinal vessels. The difficulty in extracting retinal vessels is a characteristic of retinal vessels. itself includes (central artery, central branch artery, central vein and central branch vein), the ratio of the thickness ratio between the different retinal arteries and veins (2:3), the location of the retinal artery and vein and the color. This complexity often results in errors in the retinal blood vessel extraction process, where not all blood vessel objects can be extracted properly which can reduce the accuracy of the retinal biometric system. This study will address the problem of extracting retinal vessels by proposing the use of an extraction method to produce truly unique retinal features to be included in the retinal biometric system by tracing all branches of the retinal vessels (consisting of: bifurcation, trifurcation and crossover). ). The accuracy results show that 99.81% of the images were correctly detected. The blood pattern is obtained by doing extraction which includes the preprocessing stage and is continued by doing the blood extraction stage. This pattern extraction result is used as a unique pattern to be included in the feature vector of the biometric system in identifying person based on the retina