Efficient CNN-Transformer Accelerator for Semantic Segmentation

This research focuses on optimizing hardware for computationally intensive AI tasks like semantic segmentation. The paper's contribution lies in designing a memory-compute-intensity-aware accelerator with innovative techniques like hybrid attention and cascaded pruning.

• •Focuses on hardware acceleration for semantic segmentation.
• •Employs techniques like hybrid attention and cascaded pruning for efficiency.
• •Targets energy-efficient computation with a specific technology node (28nm).