A Novel Approach to Integrated Semantic Segmentation and Feature Detection
B. Sri Ramya, Vani Pradeep, Aniruddha HD, Rahul V, S K Nithin, G V Sooraj, S Yuvaraj · 2025
This paper presents a novel approach to integrated semantic segmentation and feature detection using the You Only Look Once (YOLOv5) model. The work aims to enhance both accuracy and efficiency by unifying object detection and semantic segmentation into a single process. These tasks are often solved independently by traditional methods, which leads to a decrease in performance. With our proposed method, we are able to leverage YOLOv5's sophisticated functions in order to achieve both detection and segmentation in real-time. To assess the value of the model, experiments were conducted using the perceptual dataset, namely COCO that contains a great number of objects and complex scenes. It can be also seen that it gives approximately 3% improvement in location detection and 6% in pixel-wise segmentation over other models which proves the efficiency of this joint approach in enhancing feature detection and segmentation of elements.