Presentation Details
| AI-based applications for tracking in situ octopus body patterns and behaviors Michaella Pereira Andrade, Alan de Aguiar, João Góis, Charles Morphy D.Santos. Federal University of ABC, São Paulo, São Paulo, Brazil |
Abstract
Monitoring marine animals in their natural habitats is challenging due to underwater complexity. For octopuses, this is worsened by behavioral plasticity and the lack of standardized kinematic ethograms or databases. In this regard, there is a growing need for computational tools that tackle these barriers. We developed HideAndSeg, an AI-based tool that automates mask generation using unsupervised metrics for refinement, outperforming manual interaction models. Currently, we address octopus skin pattern identification by leveraging AI for chromatic features. AI-human classification convergence showed high performance for "Mottle" and slightly lower for other patterns. To mitigate anthropogenic bias, we shifted to unsupervised AI-based approaches. Interestingly, AI recognition prioritizes skin component relations that differ from those traditionally evaluated by human vision. We are also investigating intrinsic skin dynamics in video sequences without pre-assigned behavioral labels. This strategy extracts more replicable features, enabling the machine to identify patterns based on biological complexity that may surpass human-defined categories.
No part of this publication may be reproduced, distributed, or transmitted in any form or by any means, including photocopying, recording, or other electronic or mechanical methods, without the prior written permission of the author.
No part of this publication may be reproduced, distributed, or transmitted in any form or by any means, including photocopying, recording, or other electronic or mechanical methods, without the prior written permission of the author.
Content Locked. Log into a registered attendee account to access this presentation.