Panoptic image segmentation

Abstract

The use of Machine Learning to perform tasks that can be automated has skyrocketed due to technological advancements and greater ease of its implementation. As a result, the application of Machine Learning to carry out everyday tasks has become increasingly commonplace, such as in autonomous driving. This project focuses on the Panoptic Image Segmentation task, an image segmentation method used for Computer Vision tasks, and aims to implement a real-time danger detection system. The model used to tackle the task will be trained and evaluated on various datasets, such as the Microsoft COCO dataset and the SUN RGB-D dataset.