Computer Vision Based AutoML Platform

Abstract

The rapid increase in data generation with technological developments and scientific research has led to the development of new data analysis tools, including Machine Learning (ML). ML is an alternative to conventional engineering methods and it does not require field knowledge of the problem to obtain solutions. However, ML algorithms can be complex, and expert knowledge is required to use them effectively. Various methods have emerged to answer this problem. There are many areas and problems where machine learning can be applied. We have limited our research to the problems that can be solved by Computer Vision and AutoML. We use the AutoML and Vision based approach to achieve a solution for Classification, Object Detection and Segmentation tasks. Our aim was the develop a platform that works without expert intervention. Users loads the dataset and choose the tasks that they want and without touching anything they can train their model. When trainning has done, they can stream it in real time with their own hardware.