In the aim of natural interaction with machines, a framework must be developed to include the adaptability humans portray to understand gestures from context, from a single observation or from multiple observations. This is also referred as adaptive shot learning – the ability to adapt the mechanism of recognition to a barely seen gesture, well-known or entirely unknown. Of particular interest to the community are zero-shot and one-shot learning, given that most work has been done in the N-shot learning scenario.

Previous approaches related to zero and one-shot gesture recognition rely heavily on statistical and data mining based solutions, and leave aside the mechanisms humans use to perceive and execute gestures which can provide valuable context information. This gap leads to suboptimal solutions. This workshop aims to present and disseminate novel approaches considering the process that leads to the realization of a gesture, rather on the gesture itself. The workshop aims to encourage works that focus on the way in which humans produce gestures – the kinematic and biomechanical characteristics, and the cognitive process involved when perceiving, remembering and replicating them.

Benchmarks and performance metrics of such approaches are also of interest. For example, gesture recognition similar to that exhibited by humans are preferable to perfect recognition. In such context mimicry is more important than optimal recognition accuracy. Furthermore, how efficiency is obtained when compared to traditional N-shot learning approaches is not trivial.

We expect to bring a diverse community of linguists, psychologists, computer scientists, roboticists, and engineers together to address these questions and propose solutions to these challenges. With the workshop we expect to gain new understanding about how humans generalize from single or unseen observations from context. Knowledge gained through the workshop can shed light to understanding how infants learn to gesture, how to identify spontaneous gestures and/or uncontrolled gesturing (e.g. such as in Parkinson's disease). All these are cases of zero or one shot gesture learning or production.