DEVELOPMENT OF WEB-BASED SEMI-AUTOMATIC IMAGE LABELING TOOLS

• Main Author: Rahadian Alamsyah P W, M
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/69188
• Institution: Institut Teknologi Bandung

Computer vision has been further researched to assist human workers in various fields. The current computer vision approach uses machine learning, so the development of computer vision can take advantage of labeled image data. For the model built to process the image properly, a lot of labeled data is needed. Therefore, developing a computer vision model that utilizes machine learning has a long process of collecting and labeling image data. Many tools have been created to assist the process of labeling image data that have a graphical interface to make it easier for users to label images. In this final project, further development of the image data labeling tool is carried out to make the image data labeling process more straightforward. By utilizing a machine learning model in computer vision that has given good results in the detection of image data, the labeling tool is further modified by providing a label recommendation system when the labeling process is carried out. The recommendation system built makes the semi-automatic image data labeling process because the user does not need to provide a label from zero so that the efficiency of data labeling has the potential to increase. Kakas is built on a web-based basis, so the data integration process between labelers becomes more accessible and faster. The built-in tool can potentially increase the efficiency of labeling image data by 62.79% compared to the manual labeling processes.