Weakly-supervised Semantic Segmentation through Projective Cycle-consistency

This project involves the application of self-supervised segmentation techniques to improve the understanding of complex scenes, particularly in medical settings like surgical operating rooms. The key focus is on enabling the system to autonomously learn detailed aspects of scene understanding with minimal reliance on annotated data. This approach is especially pertinent in medical contexts where detailed and accurate scene interpretation is crucial, but annotated data may be scarce or difficult to obtain. The project’s goal is to enhance the precision and reliability of automated scene analysis in critical environments like surgeries, potentially leading to better monitoring and assistance during medical procedures.