Constructing the Equation
A scene, whether observed in person or in a frame of video, is not read as a set of isolated objects.
The simple form
Scene Intelligence is the measurement of a participant’s understanding of a scene, U, from its setting (En), its visible elements (El), the prior behaviours behind it (Pb) and the persons in it (P), for a participant, Π, who is established first.
Environment (En)
Environment (En): the setting, scenery, location, time and other contextual elements.
Elements (El)
Elements (El): the identifiable items within the scene, such as furniture, clothing, location, currency, tools, communications and more.
Prior Behaviours (Pb)
Prior Behaviours (Pb): prior actions or influences, such as the creator’s history, previous works, storyline, actors, directors, studios and producers.
Persons (P)
Persons (P): the individuals involved, one or many, whether actual people or actors, along with their history, personality and motivations.
Why weightings are required
The four factors do not contribute equally, and their relative importance moves with the context.
Each factor therefore needs a weight, and the weights are set for the query in hand.
No four-factor description captures everything a scene conveys.
A residual term, ε, is therefore added.
The equation over time
A scene is not a single frame.
The weights that fit one shot need not fit the next, and the weights that fit one era need not fit another.
The same object can carry a different weight in a different era.
Working the Z
Working the Z means working scene depth and time together, contemporaneously.
The Z must be worked equally on both sides: in the scene, and in the participant who takes it in, who also has a depth, their history, culture and beliefs, and a time, their state at that moment.
Dirichlet
Dirichlet is essential to scene intelligence as a discipline that stands on its own: scene intelligence is the science of measuring the rational against the irrational.
Behaviour that looks irrational is rational over an expanded set of contemporaneous inputs.
As the set of inputs present at the moment of the behaviour grows, the part of the behaviour left unexplained, the residual ε, shrinks.
Validity: the VITAL principle
Validity, in this sense, is not a certificate issued once.
It is the state of the weights after every test so far, and it grows with each one.
Trace and blob
The author combined trace-and-blob tracking with a library of visually based tuples, so that the next form and movement of a person can be computed from what has been tracked so far.