Here showcase material will be published as they become available.
The OPPORTUNITY dataset has been recorded to recognize complex activities in highly rich sensor environments, thus allowing to simulate opportunistic sensor configurations.
The recognition scenario is a breakfast scenario with a rich set of activity primitives, and high level activities.
The dataset contains the recordings of 72 sensors of 10 modalities distributed in 15 sensor networks.
Twelve subjects participated to the recordings, with 5 "breakfast runs" per subject, and a "drill run" dedicated to generate a lot of activity primitives (~2.5hrs/subject).
Over 11000 and 17000 object and environment interactions occurred during the recording.
Click here to see a list of publication using the Opportunity dataset.
This video shows the OPPORTUNITY Framework in four application cases: goal querying and sensor conguration, sensor appearance/disappearance, sensor learning from each other, and sensor self trust adaptation. It illustrates the dynamics of the opportunistic system as it adapts to the available sensing infrastructure at runtime.
This is a dataset from an experimental setup (gesture based HCI scenario) that allows the joint investigation of activity recognition, ErrP detection, and the combination of both into an autonomously adaptive activity recognition system.
Some characteristics of the dataset:
Do not hesitate to contact the consortium.
We develop opportunistic activity recognition systems: goal-oriented sensor assemblies spontaneously arise and self-organize to achieve a common activity and context recognition. We develop algorithms and architectures underlying context recognition in opportunistic systems.
Subscribe to our newsletter for regular project updates.