AbstractWith the advent of data management and analytics in many day life applications, recent trendsabout the various activities about « computerised restaurant » are emerging that make cooperatecomputer science and cooking or food-oriented activities.Among these activities, observing a scene, for instance a diner, and interactions among theprotagonists tend to be more and more relevant. Analysing such social interaction relies oninterpreting man-man interaction, that, from a spatial and visual point of view, relies on variousmodality: speaking, sightseeing, moving, walking, etc. with multiple cues detected in order toidentify a behaviour.Various inputs can be considered as postural congruence, left space and mutual distance, choiceof the seat, etc. Voice and sound (silence) are also relevant to this analysis: voice level, silences,onomatopoeia, etc. Voice transcription identifies spot words and can bring semantics by keywords analysing.Implementing this observation is nowadays enabled by advanced applications that embody moreand more intelligence and the corresponding interconnected, smart, embedded devices. Suchsensors, cameras, etc. make possible no intrusive captures such as video detection.Our aim is to survey how designing and implementing an observatory in order to give a contextto the main interrogation: how analysing, mining, eliciting data, crowdsourcing, video, socialnetworks, etc. can help us to observe new insights on our eating habits and choices?As we need « in vivo » capture and observation, it must offer various sceneries to avoid anypredefined scenario.