Sarah Purnamawati
Universitas Sumatera Utara

Published : 4 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 4 Documents
Search

Advertisement billboard detection and geotagging system with inductive transfer learning in deep convolutional neural network Romi Fadillah Rahmat; Dennis Dennis; Opim Salim Sitompul; Sarah Purnamawati; Rahmat Budiarto
TELKOMNIKA (Telecommunication Computing Electronics and Control) Vol 17, No 5: October 2019
Publisher : Universitas Ahmad Dahlan

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.12928/telkomnika.v17i5.11276

Abstract

In this paper, we propose an approach to detect and geotag advertisement billboard in real-time condition. Our approach is using AlexNet’s Deep Convolutional Neural Network (DCNN) as a pre-trained neural network with 1000 categories for image classification. To improve the performance of the pre-trained neural network, we retrain the network by adding more advertisement billboard images using inductive transfer learning approach. Then, we fine-tuned the output layer into advertisement billboard related categories. Furthermore, the detected advertisement billboard images will be geotagged by inserting Exif metadata into the image file. Experimental results show that the approach achieves 92.7% training accuracy for advertisement billboard detection, while for overall testing results it will give 71,86% testing accuracy.
Astrocytoma, ependymoma, and oligodendroglioma classification with deep convolutional neural network Romi Fadillah Rahmat; Mhd Faris Pratama; Sarah Purnamawati; Sharfina Faza; Arif Ridho Lubis; Al-Khowarizmi Al-Khowarizmi; Muharman Lubis
IAES International Journal of Artificial Intelligence (IJ-AI) Vol 11, No 4: December 2022
Publisher : Institute of Advanced Engineering and Science

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.11591/ijai.v11.i4.pp%p

Abstract

Glioma as one of the most common types of brain tumor in the world has three different classes based on its cell types. They are astrocytoma, ependymoma, oligodendroglioma, each has different characteristics depending on the location and malignance level. Radiological examination by medical personnel is still carried out manually using magnetic resonance imaging (MRI) medical imaging. Brain structure, size, and various forms of tumors increase the level of difficulty in classifying gliomas. It is advisable to apply a method that can conduct gliomas classification through medical images. The proposed methods were proposed for this study using deep convolutional neural network (DCNN) for classification with k-means segmentation and contrast enhancement. The results show the effectiveness of the proposed methods with an accuracy of 95.5%.
Android-based automatic detection and measurement system of highway billboard for tax calculation in indonesia Romi Fadillah Rahmat; Sarah Purnamawati; Handra Saito; Muhammad Fariz Ichwan; Tri Murti Lubis
Indonesian Journal of Electrical Engineering and Computer Science Vol 14, No 2: May 2019
Publisher : Institute of Advanced Engineering and Science

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.11591/ijeecs.v14.i2.pp877-886

Abstract

Billboards are objects, tools or actions, which based on the characteristics serve its own purpose to earn profits, advertise certain people or service, and to draw public’s attention by placing it in a very strategic place. It has led the government to charge tax on billboards based on its location, dimensions, and viewpoints. Therefore, authorized parties have to be able to ensure the data authenticity of the proposed billboards. One of the obstacles in data verification is the time of billboards measurement process due to its size and height from the ground, based on this problem, and we developed a system which can measure the dimensions of billboards without physically touching it by implementing image processing methods to identify the billboards. The implementation is by measuring the dimensions of the billboards using perspective concept, then calculates the distance between the camera and the object using two-point distance calculation GPS coordinates. The results showed that the distance calculation using the GPS method generated inaccurate values, whereas the systematic distance method generated a result of errors’ range from 0.5 to 25 cm if the image acquisition is performed nearly perpendicular to the object.
Vacant parking space identification using probabilistic neural network Romi Fadillah Rahmat; Sarah Purnamawati; Joko Kurnianto; Sharfina Faza; Muhammad Fermi Pasha
Indonesian Journal of Electrical Engineering and Computer Science Vol 14, No 2: May 2019
Publisher : Institute of Advanced Engineering and Science

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.11591/ijeecs.v14.i2.pp887-894

Abstract

The need for public parking space is increasing nowadays due to the high number of cars available.  Users of car parking services, in general, are still looking for vacant parking locations to park their vehicle manually. With the current technological developments, especially in image processing field, it is expected to solve the parking space problem. Therefore, this research implements image processing to determine the location of vacant parking space or occupied ones that run in real-time. In this study, the proposed method is divided into five stages. The first stage is image acquisition to capture the image of parking location. Then it continues to pre-processing stage which consists of the process of saturation, grayscale and thresholding. The third stage is image segmentation to cut the image into five parts. The next stage is feature extraction using invariant moment, and the last stage would be identification process to determine the location of vacant parking spaces or occupied ones. The results of this research using 100 test images generates an accuracy, recall, and precision of 94%.