Towards ontology driven learning of visual concept detectors
This work addresses the problem of arbitrary concept detection in videos for users needing semantic search capabilities, representing an incremental improvement by integrating existing components.
The paper tackles the challenge of detecting arbitrary concepts in arbitrary videos by proposing a system that combines neural techniques, a large-scale visual concepts ontology, and an active learning loop for on-the-fly model learning, resulting in improved recall of concept detection and enabling semantic search on annotated video libraries.
The maturity of deep learning techniques has led in recent years to a breakthrough in object recognition in visual media. While for some specific benchmarks, neural techniques seem to match if not outperform human judgement, challenges are still open for detecting arbitrary concepts in arbitrary videos. In this paper, we propose a system that combines neural techniques, a large scale visual concepts ontology, and an active learning loop, to provide on the fly model learning of arbitrary concepts. We give an overview of the system as a whole, and focus on the central role of the ontology for guiding and bootstrapping the learning of new concepts, improving the recall of concept detection, and, on the user end, providing semantic search on a library of annotated videos.