Spatio-Temporal Degree of Hoarding Clutter Using SAM and Canny Edge Detection Against LMM

The previous paper proposed the SAM-HOA (SAM-based degree of HOArding clutter in Anywhere) to be time-sequentially monitored with only an ordinary camera. This paper conducts a novel study for developing a spatio-temporal degree of hoarding clutter using not only SAM (Segment Anything Model) but also Canny edge detection to improve its performance and robustness, and some experiments to validate the improved metric. In addition, this paper also conducts a study for developing a comprehensive management system, SAM-HOACS, for a user with/out her/his family member(s) and pet animal(s) between her/his biosensor information and each degree of hoarding clutter for her/his personal room and/or shared room as her/his domestic behaviors. The future work by the proposed system would reveal new findings about relationships between the improved SAM-HOA and biosensor information (e.g., quality of sleep, daytime activity), its subjects (e.g., family members, pet animals) or contexts (e.g., day of week, events).