Object Instance Identification in Dynamic Environments
This work addresses object instance identification for applications in dynamic settings like human-object interaction, but it is incremental as it focuses on analysis and benchmark creation without proposing a new method.
The paper tackles the problem of identifying object instances in dynamic environments where appearance changes due to interactions, occlusion, and background shifts, leading to high intra-instance variation. It introduces a new benchmark of over 1,500 instances based on EPIC-KITCHENS and identifies key challenges for improvement, such as robustness to appearance changes and feature integration.
We study the problem of identifying object instances in a dynamic environment where people interact with the objects. In such an environment, objects' appearance changes dynamically by interaction with other entities, occlusion by hands, background change, etc. This leads to a larger intra-instance variation of appearance than in static environments. To discover the challenges in this setting, we newly built a benchmark of more than 1,500 instances built on the EPIC-KITCHENS dataset which includes natural activities and conducted an extensive analysis of it. Experimental results suggest that (i) robustness against instance-specific appearance change (ii) integration of low-level (e.g., color, texture) and high-level (e.g., object category) features (iii) foreground feature selection on overlapping objects are required for further improvement.