A Deep CNN Model Based Ensemble Approach for Semantic and Instance Segmentation of Indoor Environment

Image segmentation has emerged as a popular and challenging task in computer vision. Deep learning methods have demonstrated significant success in this domain. In this paper, we propose a deep ensemble model for semantic segmentation of indoor scenes. Our ensemble model combines three distinct deep learning models to enhance both diversity and accuracy. Additionally, we customize YOLOv8 for instance-level segmentation of indoor environments. A comprehensive set of experiments on the publicly available data set demonstrates that our model achieves the highest IoU compared to state-of-the-art deep learning models.