Gary A. Atkinson

About

Gary A. Atkinson is a researcher whose work bridges computer vision, robotics, and deep learning, with a particular focus on challenging real-world environments. His key contributions center on developing advanced image segmentation techniques for complex, low-visibility scenes, such as underfloor and subterranean spaces. Atkinson is best known for his pioneering application of mask region-based convolutional neural networks (Mask R-CNN) to these domains, demonstrating how two-stage transfer learning can dramatically improve segmentation accuracy in data-scarce settings. His most cited paper, "Image segmentation of underfloor scenes using a mask regions convolutional neural network with two-stage transfer learning" (2020, 35 citations), exemplifies this approach, offering a robust solution for automated inspection in construction and infrastructure maintenance. This work has significant implications for reducing manual inspection costs and enhancing safety in hazardous environments. Atkinson’s research is notable for its practical impact, integrating state-of-the-art AI with domain-specific challenges, and his findings continue to inform the development of autonomous systems for industrial and robotic applications.

Research Focus

Key Achievements

Top Papers

1. 1
   Image segmentation of underfloor scenes using a mask regions convolutional neural network with two-stage transfer learning

   35 citations · 2020

Key Collaborators

Contact & Links