Context-aware classification of continuous video from wearables

Abstract

Name and full contact address (surface, fax, email) of the individual responsible for submitting and receiving inquiries about the submission: Contact Lutz Dickmann, AbstractThe next big step in video-based life logging is to exploit image processing and context inference from multiple sensors. Automatic segmentation and classification of personal experiences recorded with always-on wearable devices may forge entirely new pragmatics of human interaction. Yet, how do we anticipate social dynamics to tackle critical ethical matters? Are we supposed to fit a design concept onto a to this point non-existent market or technology? We aim to contribute to the digital age discourse by making this emergent domain viable for responsible iterative design approaches. This sketch outlines how we address the novel technology at hand by integrating technical development and design strategies in a convergent full-circle approach.

A Framework for Review, Annotation, and Classification of Continuous Video in Context

“…The interdisciplinary origin of our software development efforts is elaborated on in another publication. [12] Our Cocoa-based implementation adheres to the model/view/controller paradigm and is written in Objective-C/C++, which allows seamless integration of libraries in the ANSI C or C++ dialects, in our case most notably OpenCV (for feature extraction and analysis), FFmpeg (for video decoding), and the C Clustering Library (for efficient clustering). Python bindings allow quick experiments with a wide range of algorithms from the Pythonfriendly scientific computing community.…”