Table 2.

Challenges and research opportunities in adopting eating action unit–based sensing methods.

Challenge	Research opportunity
Limited understanding of context surrounding system success and failures	Use wearable video cameras to validate contextual information surrounding when sensor–algorithm pairings fail in real-world settings
Privacy protection in ground truth data collection methods	Identify novel ways of protecting bystanders and other sensitive information in the field of view of cameras both in hardware and software to ensure wearer privacy concerns are addressed, thereby increasing likelihood of capturing naturally occurring behavior
Inability to accurately distinguish between food categories	Define food categories that are most useful for clinicians and researchers for diet interventions and food recalls