Dolphin picture

UBISea the next step

Almost a year ago we presented you UbiSea, a prototype of a connected buoy, what happened to it? The answer is in the following article!

Ubidreams
Share:
  • You have not accepted cookies to enable this service. You can change your cookies.
  • You have not accepted cookies to enable this service. You can change your cookies.
  • You have not accepted cookies to enable this service. You can change your cookies.

The UBISea project

The UBISea project is jointly led by Ubidreams, Systel and the L3i Laboratory in La Rochelle. Its objective is to prototype a network of connected buoys capable of detecting and monitoring in real time acoustic or visual phenomena in a marine environment. One of the applications of this tool is to provide local authorities with the means to monitor specific areas and to alert in case of fraudulent presence or behaviour.

After a first feasibility study focused on ship detection, a new part of the project aims to adapt the device to the detection of marine mammals (Common Dolphin) in marine areas impacted by fishing activity. Such a device would provide assistance for local management of the marine environment, by allowing the marking of zones of exclusion of activity that are more precise in geographical and temporal terms, and closer to real needs. 

Artificial intelligence for marine environment management

The next step of the project is the automated acoustic recognition of clicks and whistles of common dolphins, using artificial intelligence via the training of a deep neural network (Deep Learning). This work will be carried out in partnership with the Pelagis Observatory [https://www.observatoire-pelagis.cnrs.fr/] in La Rochelle, which specialises in the observation and study of marine mammals and birds. Their expertise will enable them to provide the labelled acoustic data required for training the model.

The project will first test different neural networks, using several technologies (Ketos [https://docs.meridian.cs.dal.ca/ketos/] package implemented on Tensorflow, Pytorch), and then compare their efficiency and portability on an autonomous embedded system.

DevelopmentIOT

A question, a project ? We help you define your project.