Novel event analysis for human-machine collaborative underwater exploration

Highlights：

• The major novelty and contribution of this work is that we propose a new problem, i.e., novel deep sea event analysis for deep sea scientific exploration. To our best knowledge, this is the first work to handle deep sea extreme event analysis with multimedia technologies, including novel event detection, tracking and summarization simultaneously.

• For some key models used by work previous work, we propose a new online novel event tracking to overcome non-rigid deformation and a visual saliency detection via simple structured deep learning for novel event detection in this revised paper. We mainly concern the issue of the power / consumption resource limitation and design a general low cost visual framework including three components.

• Moreover, we also collect and build a new deep sea novel event visual dataset, which is totally new and we will try to release it soon.