Utilizing YoLov8 Architecture for Sharp-Edge Segmentation

Abstract

In computer vision applications, such as object identification, scene comprehension, and robotic manipulation, accurate object segmentation is essential. In complicated contexts, traditional segmentation techniques frequently have trouble distinguishing sharp objects with accuracy. This work declare a novel approach to robust and exact object segmentation utilizing deep learning techniques built around the YOLOv8 architecture using an annotated collection of sharp object photos, the suggested methodology entails pre-processing the images with the CVAT(Computer Vision Annotation Tool) tool to create binary masks. The YOLOv8 model is then trained on these annotated photos, and its behavior is compared with conventional segmentation techniques using a various image datasets. With its potential to advance the realm of computer vision, the YOLOv8 architecture presents a promising option for precise and effective sharp object segmentation in real-world scenarios.