Jurnal RESTI (Rekayasa Sistem dan Teknologi Informasi)
Not a member yet
1071 research outputs found
Sort by
Use of Plant Health Level Based on Random Forest Algorithm for Agricultural Drone Target Points
Get PDFChemical residues from the use of pesticides in agriculture can impact human health through environmental and food pollution. To lessen the negative effects of excessive pesticide use, pesticides must be applied to plants by dose. The dose of pesticide application can be based on a plant health level, which is the result of drone Normalized Difference Vegetation Index (NDVI) image analysis. Drones can also be used for spraying pesticides. Analysis of plant health levels was carried out using the Random Forest (RF) algorithm. The results of the classification plant health levels will be used to design spray drone flight routes. The objective of this research is to classify plant health levels of rice based on NDVI imagery using the RF algorithm and to compile a database of spray drone target points. The results of this study indicate that the classification of plant health levels using the RF algorithm produces an accuracy value of 98% and a Kappa value of 0.96. As a result, the model developed and the algorithm employed is quite effective at classifying the level of plant health. Furthermore, spray drone target points based on plant health levels can be generated. Optimally the spray distance between rows is 2 m.
Chemical residues from the use of pesticides in agriculture can impact human health through environmental and food pollution. To lessen the negative effects of excessive pesticide use, pesticides must be applied to plants by dose. The dose of pesticide application can be based on a plant health level, which is the result of drone Normalized Difference Vegetation Index (NDVI) image analysis. Drones can also be used for spraying pesticides. Analysis of plant health levels was carried out using the Random Forest (RF) algorithm. The results of the classification plant health levels will be used to design spray drone flight routes. The objective of this research is to classify plant health levels of rice based on NDVI imagery using the RF algorithm and to compile a database of spray drone target points. The results of this study indicate that the classification of plant health levels using the RF algorithm produces an accuracy value of 98% and a Kappa value of 0.96. As a result, the model developed and the algorithm employed is quite effective at classifying the level of plant health. Furthermore, spray drone target points based on plant health levels can be generated. Optimally the spray distance between rows is 2 m
Game Design for Mobile App-Based IoT Introduction Education in STEM Learning
Get PDFSTEM education has received considerable attention in recent years. However, developing valid and reliable assessments in interdisciplinary learning in STEM has been a challenge. Therefore, many students ranging from junior high school to university students are only familiar with the Internet of Things (IoT) from social media but do not know its concept and function in STEM learning. This is also supported by the absence of educational applications about IoT. This research aims to introduce IoT by using mobile applications. This research refers to the multimedia development method according. The data collection method in this study was carried out by means of observation and interviews randomly to high school students to university students. This data collection was carried out using the experimental method of application testing to analyze user needs from several aspects such as features, images, and fonts. This research is also supported by the existence of literature studies derived from several journals. The results show that the functions in the application can operate as expected. Based on the survey results of the application, 75.37% of respondents rated this application in the very good category and gave positive responses so that this application could be well received by users
STEM education has received considerable attention in recent years. However, developing valid and reliable assessments in interdisciplinary learning in STEM has been a challenge. Therefore, many students ranging from junior high school to university students are only familiar with the Internet of Things (IoT) from social media but do not know its concept and function in STEM learning. This is also supported by the absence of educational applications about IoT. This research aims to introduce IoT by using mobile applications. This research refers to the multimedia development method according. The data collection method in this study was carried out by means of observation and interviews randomly to high school students to university students. This data collection was carried out using the experimental method of application testing to analyze user needs from several aspects such as features, images, and fonts. This research is also supported by the existence of literature studies derived from several journals. The results show that the functions in the application can operate as expected. Based on the survey results of the application, 75.37% of respondents rated this application in the very good category and gave positive responses so that this application could be well received by user
Application of Deep Convolutional Generative Adversarial Networks to Generate Pose Invariant Facial Image Synthesis Data
Get PDFThe field of technology is currently developing rapidly, one of the developments is artificial intelligence. Artificial intelligence can still find it difficult to solve problems that are easy for humans to do but difficult for computers to describe, such as facial recognition. There are still several problems related to the existing facial recognition model, namely, the facial recognition model is still unable to recognize facial shapes that are not in a perfect state due to several factors such as face position, lighting, expression, and obstacles covering the face. Among these several factors, the most influencing factor is the position of the face. Therefore, in this study, deep convolutional generative adversarial networks (DCGANs) will be applied to generate fake image data with varying face positions. This research will be carried out starting from collecting data, processing data, designing and training models, hyperparameter tuning, and lastly analyzing test results. Based on the results of hyperparameter tuning that were performed sequentially, the best hyperparameter combination produced is 200 epoch, 0.002 Generator learning rate, 0.5 Generator momentum/beta1, Adam as Generator optimizer, 0.0002 Discriminator learning rate, 0.5 Discriminator momentum/beta1, and Adam as Discriminator optimizer. The combination of hyperparameters gives a result with an FID score of 74.05. Based on testing with human observers, generated fake images have relatively good results, but there are still few bad fake image result
Analysis of Bit Rate and Distance Variation on Multiplexing System of Indoor Li-Fi Technology Using Movable LED Panel
Get PDFThe major problem of using light fidelity (Li-Fi) technology is still limited to the line of sight (LOS) conditions, which poses a challenge to perform bandwidth efficiency to support increased bit rates, especially for indoor use. In addition, the distance between the lamp driver (transmitter) and the receiver becomes a critical discussion to determine the characteristics of propagation losses. Therefore, this study focuses on analyzing the performance of an indoor Li-Fi multiplexing system using a movable LED panel (LP) based on parameters of bit rate and distance variation on multiple-input multiple-output (MIMO) 2x2 and 4x4. The parameter analysis of signal quality included optical and electrical signal spectrum characteristics, signal-to-noise ratio (SNR), bit error rate (BER), and Q-factor parameters. Based on the results, the increase in bit rate and distance significantly increases the BER value and decreases the Q-factor value. Both the 2x2 and 4x4 mux systems can meet standards up to a bit rate of 30 Mbps at a LOS distance of 3 meters, while at a bit rate of 40 Mbps, there are no channels that meet the ITU-T standard. In addition, the quality of the signal received at a distance of 4 meters, the 2x2 mux system can only reach the standard at a bit rate of 20 Mbps for all channels. However, channel 3 and channel 4 on the 4x4 mux system model still have the BER and Q-factor values that meet the standard in the bit rate of 30 Mbps. However, the decrease in the SNR value affected by the bit rate increase and distance is insignificant. Therefore, it becomes an opportunity for further observation of the proposed multiplexing system, detection scheme, or responsivity, and signal processing on the receiver side to be reliable on the higher bit rate.The major problem of using light fidelity (Li-Fi) technology is still limited to the line of sight (LOS) conditions, which poses a challenge to perform bandwidth efficiency to support increased bit rates, especially for indoor use. In addition, the distance between the lamp driver (transmitter) and the receiver becomes a critical discussion to determine the characteristics of propagation losses. Therefore, this study focuses on analyzing the performance of an indoor Li-Fi multiplexing system using a movable LED panel (LP) based on parameters of bit rate and distance variation on multiple-input multiple-output (MIMO) 2x2 and 4x4. The parameter analysis of signal quality included optical and electrical signal spectrum characteristics, signal-to-noise ratio (SNR), bit error rate (BER), and Q-factor parameters. Based on the results, the increase in bit rate and distance significantly increases the BER value and decreases the Q-factor value. Both the 2x2 and 4x4 mux systems can meet standards up to a bit rate of 30 Mbps at a LOS distance of 3 meters, while at a bit rate of 40 Mbps, there are no channels that meet the ITU-T standard. In addition, the quality of the signal received at a distance of 4 meters, the 2x2 mux system can only reach the standard at a bit rate of 20 Mbps for all channels. However, channel 3 and channel 4 on the 4x4 mux system model still have the BER and Q-factor values that meet the standard in the bit rate of 30 Mbps. However, the decrease in the SNR value affected by the bit rate increase and distance is insignificant. Therefore, it becomes an opportunity for further observation of the proposed multiplexing system, detection scheme, or responsivity, and signal processing on the receiver side to be reliable on the higher bit rate
MobileNetV3-based Handwritten Chinese Recognition Towards the Effectiveness of Learning Hanzi
Get PDFWriting Mandarin characters is considered the most challenging component for beginners due to the rules and character formations. This paper explores the potential of a machine learning-based digital learning tool for writing Mandarin characters. It also conducts a comparative study between MobileNetV2 and MobileNetV3, exploring different configurations. The research follows the Multimedia Development Life Cycle (MDLC) method to create both the application and machine learning models. Participants from higher education institutions that offer Mandarin courses in Batam, Indonesia, were involved in a User Acceptance Test (UAT). Data was gathered through questionnaires and analyzed using the System Usability Scale (SUS) methods. The results show positive user acceptance, with an SUS score of 77.92%, indicating a high level of acceptability. MobileNetV3Small was also preferred for recognizing the user’s handwriting, due to comparable accuracy size, rapid inference time, and smallest model size. While the application was well-received, several participants provided constructive feedback, suggesting potential improvements
Comparison of the Accuracy of Drug User Classification Models Using Machine Learning Methods
Get PDFDrug abuse are on the rise, with many users enter the addiction phase, often resulting in overdose and death. Drugs are chemical compounds that are capable of affecting biological functions, and they can induce feelings of happiness and reduce pain. To address this growing problem, a proactive measure is needed. Therefore, this study aims to classify drug users and non-users, so that health workers and therapists can educate about the dangers of drugs to non-users and rehabilitate drug users. This study uses drug consumption data taken from the UCI Irvine Machine Learning Repository. The data consist of 1885 rows with 32 attributes and 2 classes, where there are 18 types of legal and illegal drugs. This research utilizes machine learning methods, specifically Artificial Neural Networks (ANN), Decision Tree (DT), K-Nearest Neighbor (KNN), Support Vector Machine (SVM) and Random Forest (RF), in addition to evaluation methods such as Confusion Matrix and Area Under Curve (AUC). The results showed that RF outperformed the other methods, with accuracy, precision, and recall of 93%, and an f1 score of 89%, while the AUC value was still suboptimal at 0.66. DT had the worst results, with 82% precision, 87% precision, 82% recall, 84% f1 score, and an AUC value of 0.56. With these results, this research can be continued into an application that can classify drug users and nonusers
Android Application for Tomato Leaf Disease Prediction Based on MobileNet Fine-tuning
Get PDFTomato is one of the most well-known and widely cultivated plants in the world. The result of tomato production is affected by the conditions of the plants when they are grown. It may decrease due to leaf plant disease caused by climate change, pollinator decrease, microbial pets, or parasites. To prevent this, an image-based application is needed to identify tomato plant disease based on visually unique patterns or marks seen on leaves. In this paper, we proposed a CNN fine-tuned model based on MobileNet architectures to identify tomato leaf disease for mobile applications. Based on the results tested by K-fold cross-validation, the best accuracy achieved by the proposed model is 97.1%. Additionally, the best average precision, recall and F1 Score are 99.8%, 99.8%, and 99.5%, respectively. The model with the best results is also implemented into Android-based mobile applications
Digital Image Encryption Using Logistic Map
No full textThis study focuses on the application of the logistic map algorithm in the Python programming language for digital image encryption and decryption. It investigates the impact of image type, image size, and logistic map parameter values on computational speed, memory usage, encryption, and decryption results. Three image sizes (300px 300px, 500px x 500px, and 1024px x 1024px) are considered in TIFF, JPG, and PNG formats. The digital image encryption and Decryption process utilizes the logistic map algorithm implemented in Python. Various parameter values are tested for each image type and size to analyze encryption and decryption outcomes. The findings indicate that the type of image does not affect memory usage, which remains consistent regardless of image type. However, image type significantly influences the decryption results and computation time. In particular, the TIFF image type exhibits the fastest computation time, with durations of 0.17188 seconds, 0.28125 seconds, and 1.10938 seconds for 300px x 300px, 500px x 500px, and 1024px x 1024px images, respectively. In addition, the encryption results vary depending on the type of image. The logistic map algorithm is unable to restore encryption results accurately for JPG images. Furthermore, research highlights that higher values of x, Mu and Chaos lead to narrower histogram values, resulting in improved encryption outcomes. This study contributes to the field by exploring the application of the logistic map algorithm in Python and analyzing the effects of image type, image size, and Logistic Map parameter values on computation time, memory usage, and digital image encryption and Decryption results. Prior research has not extensively addressed these aspects in relation to the Logistic Map algorithm in Python
Modification of SqueezeNet for Devices with Limited Computational Resources
No full textIn recent years, the computational approach has shifted from a statistical basis to deep neural network architectures which process the input without explicit knowledge that underlies the model. Many models with high accuracy have been proposed by training the datasets using high performance computing devices. However, only a few studies have examined its use on non-high-performance computers. In fact, most users, who are mostly researchers in certain fields (medical, geography, economics, etc.) sometimes need computers with limited computational resources to process datasets, from notebooks, personal computers, to mobile processor-based devices. This study proposes a basic model with good accuracy and can run lightly on the average computer so that it remains lightweight when used as a basis for advanced deep neural networks models, e.g., U-Net, SegNet, PSPNet, DeepLab, etc. Using several well-known basic methods as a baseline (SqueezeNet, ShuffleNet, GoogleNet, MobileNetV2, and ResNet), a model combining SqueezeNet with ResNet, termed Res-SqueezeNet, was formed. Testing results show that the proposed method has accuracy and inference time of 84.59% and 8.46 second, respectively, which has an accuracy of 2% higher than the SqueezeNet (82.53%) and is close to the accuracy of other baseline methods (from 84.93% to 0.88.01%) while still maintaining the inference speed (below nine second). In addition, residual part of the proposed method can be used to avoid vanishing gradient, hence, it can be implemented to solve more advanced problems which need a lot of layers, e.g., semantic segmentation, time-series prediction, etc.
In recent years, the computational approach has shifted from a statistical basis to deep neural network architectures which process the input without explicit knowledge that underlies the model. Many models with high accuracy have been proposed by training the datasets using high performance computing devices. However, only a few studies have examined its use on non-high-performance computers. In fact, most users, who are mostly researchers in certain fields (medical, geography, economics, etc.) sometimes need computers with limited computational resources to process datasets, from notebooks, personal computers, to mobile processor-based devices. This study proposes a basic model with good accuracy and can run lightly on the average computer so that it remains lightweight when used as a basis for advanced deep neural networks models, e.g., U-Net, SegNet, PSPNet, DeepLab, etc. Using several well-known basic methods as a baseline (SqueezeNet, ShuffleNet, GoogleNet, MobileNetV2, and ResNet), a model combining SqueezeNet with ResNet, termed Res-SqueezeNet, was formed. Testing results show that the proposed method has accuracy and inference time of 84.59% and 8.46 second, respectively, which has an accuracy of 2% higher than the SqueezeNet (82.53%) and is close to the accuracy of other baseline methods (from 84.93% to 0.88.01%) while still maintaining the inference speed (below nine second). In addition, residual part of the proposed method can be used to avoid vanishing gradient, hence, it can be implemented to solve more advanced problems which need a lot of layers, e.g., semantic segmentation, time-series prediction, etc
Comparison of Mycobacterium Tuberculosis Image Detection Accuracy Using CNN and Combination CNN-KNN
No full textMycobacterium tuberculosis is a pathogenic bacterium that causes respiratory tract disease in the lungs, namely tuberculosis (TB). The problem is to find out the bacterial colonies when the observation is still done manually using a microscope with a magnification of 1000 times. It took a long time and was tiring for the observer's eye. Based on this background, an automatic detection system for Mycobacterium tuberculosis was designed. Mycobacterium tuberculosis image data were obtained from the Semarang City Health Center. The dataset used is 220 sputum images, which are divided into 180 training data and 40 testing data. The method used in this research is a combination of Convolutional Neural Network (CNN) and K-Nearest Neighbor (KNN). CNN is used for image feature extraction. Furthermore, the results of the CNN feature extraction are classified using the KNN. The results of the accuracy of the combination of CNN-KNN and CNN were also compared. The stages of the process are color transformation, feature extraction, and data training with CNN, then classification with KNN. The results of the classification test between CNN and the CNN-KNN combination show that the CNN-KNN combination is better. The result of CNN-KNN accuracy is 92.5%, while CNN's accuracy is 90%.Mycobacterium tuberculosis is a pathogenic bacterium that causes respiratory tract disease in the lungs, namely tuberculosis (TB). The problem is to find out the bacterial colonies when the observation is still done manually using a microscope with a magnification of 1000 times. It took a long time and was tiring for the observer's eye. Based on this background, an automatic detection system for Mycobacterium tuberculosis was designed. Mycobacterium tuberculosis image data were obtained from the Semarang City Health Center. The dataset used is 220 sputum images, which are divided into 180 training data and 40 testing data. The method used in this research is a combination of Convolutional Neural Network (CNN) and K-Nearest Neighbor (KNN). CNN is used for image feature extraction. Furthermore, the results of the CNN feature extraction are classified using the KNN. The results of the accuracy of the combination of CNN-KNN and CNN were also compared. The stages of the process are color transformation, feature extraction, and data training with CNN, then classification with KNN. The results of the classification test between CNN and the CNN-KNN combination show that the CNN-KNN combination is better. The result of CNN-KNN accuracy is 92.5%, while CNN's accuracy is 90%