Extending Affordance Segmentation for Wearable Robots to Multi-Object Images
Experimental results confirm that the proposed pipeline improves target disambiguation accuracy by up to 9% compared with existing solutions, while preserving the computational cost.