Convolutional neural network (CNN) is one of the neural networks used in image data. CNN has a good ability to detect objects in an image. This study discusses the comparison of two deep learning models based on convolutional neural network, namely the Inception-V3 method and the MobileNet method. Both algorithms are analyzed fairly on gender classification using eye images. There have been many research completions that have conducted studies on gender classification based on faces, but gender classification based on eyes has many challenges. This gender classification is grouped into two classes, namely male and female. This study aims to build a gender classification model from eye image. The processes in this research include selecting the dataset, preprocessing the data, dividing the data which is divided into training data and test data, modeling, and evaluating the performance of the model. This study uses a public dataset, where the data contains a total of 2,681 images consisting of 1251 male eyes and 1430 female eyes. This study concludes that gender classification using eye image using the Inception-V3 method is better than the MobileNet method. This is obtained based on the accuracy value generated by the Inception-V3 method which is higher than the MobileNet-V2 method which obtains an accuracy of 91.82%.

A website is typically used as a medium for open, quick, and widespread information dissemination. Additionally, the website has been used for competition-related activities sponsored by an organization, such as information portal websites, registration portals, and competition evaluation media. One of the elements that determines how reliable a website is is its capacity to respond to and handle user requests. Additionally, a website that handles some information related to national competitions needs to be highly reliable. Performance testing was used in this study to evaluate how well the LO KREATIF website responded to and served users, particularly at the same time. The JMeter tool was used to conduct the performance test. The test results show that some web pages in general can serve up to 500 users at the same time stably without errors.

No abstract