Enhancing Semantic Segmentation with NMSS Model

Created using ChatSlide

The presentation introduces NMSS, a novel approach for semantic segmentation that balances accuracy and efficiency through a hybrid encoder-decoder model. Key contributions include the Natten-Mamba Block for improved feature processing and Multi-scale Feature Extraction for superior performance. Detailed methodology covers architectural designs, followed by experiments using various datasets such as ADE20K and Cityscapes. Evaluation metrics focus on the Mean Intersection over Union,...