Sediment grains
Image:

Sediment grains

Transforming ocean surveying by the power of DL and statistical methods

New project associated with Visual Intelligence on reduce bottlenecks related to uncertainty estimates in classification by a combined statistical and machine learning approach. The project is on recognition of specific objects at and within the seabed from several image types.

New associated project in Visual Intelligence: Transforming ocean surveying by the power of DL and statistical methods

The project is a collaboration between Department of Mathematics and Statistics (IMS host) and Department of Geosciences (IG) at UiT - the Arctic university of Norway.

Project Leader at IMS: Prof. Fred Godtlibsen

PIs at IG:  Reseacher  Steffen Aagaard Sørensen,  prof. Morten Hald

Funding: Research council of Norway, through the IKTPLUSS program and collaborating companies Multiconsult   and Argeo.

Summary:

The amount of data collected is rapidly increasing.  Artificial intelligence is key to extract crucial information from huge amounts of data. Classifiers suggests a class affiliation based on the input data. This problem domain is denoted classification, and machine learning and statistical techniques have proven very useful. At present, these two cultures have only to a limited extent, be able to take out synergies.

Exploring the synergies are the next step to move forward. The focus of this project is to join their strengths to get rid of bottlenecks within classification in particular related to uncertainty estimation, transferable methodology and highly heterogenous data sets.

As test cases, we focus on specific tasks that are important for both the collaboration companies and for the geoscientific community. In particular, we aim to develop a product that automatize the recognition of specific objects at and within the seabed from several image acquisition types (examples in figure below) .Moreover, we will demonstrate how a time-consuming manual procedure can be replaced by an automatic system for identification and classification of objects carrying information about climatology and environmental conditions.Successful algorithms developed in the present project will pave the way for a fully automatic system, so that the human interaction of the whole process, at some point in the future, largely can be omitted.

Examples of elements at the seabed intended for targeted automated classification: A)Large scale man-object (shipwreck) detected at the seabed via high resolution synthetic aperture sonar (HISAS), (B and C) Two video still images that shows pieces of slow growing pink Coralline algae that are difficult to distinguish,(B) Image showing the nature/habitat type «maerl bed» consisting of detached spherical algae suspected of high susceptibility to ocean acidification and whose prevalence is unknown, (C) Image showing the nature/habitat main type«Grunn marin fastbunn (M1)» where common red algae form a crust on rocks/shells etc. In areas where these two nature types mix the likeness of the red algae make identification of species and establishment of nature type boundaries difficult, (D) Sample showing sediment grains, calcareous benthic foraminifera(b) and planktic foraminifera (p) as viewed through the microscope, (E)Microplastics as viewed through the microscope and zoom in on elongated microplastic fibers (m).

 

For more information please contact: Fred Godtliebsen

Steffen Aagaard Sørensen or Morten Hald

Latest news

Pitch-day for ideas for master’s projects for the computer science and machine learning students at UiT

April 19, 2022

We invited companies, the public sector, and other organizations to present their ideas for master’s projects and to connect with students having an interest and background in computer science and machine learning.

Talk about Ethics in AI at the Aim North 2022 conference

April 6, 2022

Director Robert Jenssen gave a talk about explainable AI connected to the overarching theme of the Aim North conference, namely Ethics in AI. The talk outlined e.g. VI's recent results on detecting artifacts in medical images by our "Prototypical Relevance Propagation" method.

Talk by VI's Director Robert Jenssen at the opening of the Pioneer AI Centre

March 29, 2022

The Pioneer AI Centre is Denmark's flagship centre on AI, headed by prof. Serge Belongie, University of Copenhagen/KU. The official opening of the centre took place March 28 2022 at the "Black Diamond" in Copenhagen city centre.

Visual Intelligence is a central partner in NORA research school - a new national PhD School

March 25, 2022

The Visual Intelligence Graduate School (VIGS) will be an important component in the NORA research school and Visual Intelligence's Northern Lights Deep Learning (NLDL) conference will be a venue for winter school activities within this nation-wide initiative.

SFI Visual Intelligence guest talk by Cristina Soguero-Ruiz

March 22, 2022

Associate Professor Cristina Soguero-Ruiz visited Tromsø and gave a very interesting talk on March 17th 2022 on how to exploit AI and machine learning for medical image analysis and healthcare applications.

New associated project on offshore wind energy

February 25, 2022

Through the ongoing Akademia Avtale fund between Equinor and UiT. The goal is to develop new innovative approaches for combining image data and other sources of information in the context of wind energy.

Program for Visual Intelligence Days 2021

October 15, 2021

The "Visual Intelligence days" is coming up shortly, and VI-collaborators will gather in Oslo to connect and share knowledge. Phd-students and early career researchers will also gather one day in advance to socialize. Please see the updated program.

How is AI affecting democratic processes?

October 4, 2021

Director Robert Jenssen gave a talk and participated in a panel debate organized by the Norwegian Academy of Science and Letters on the topic “How is AI affecting democratic processes?“.

Machine learning talk at University of Bristol Summer school

September 8, 2021

Director of Visual Intelligence Robert Jenssen gave a talk at the BIAS Summer school at University of Bristol. The talk describes how deep learning is leveraged in computer vision in an industrial innovation project for monitoring power lines in Norway.

Paper published in International Journal of Remote Sensing

June 28, 2021

Qinghui Liu, Michael Kampffmeyer, Robert Jenssen and Arnt-Børre Salberg have published their paper Self-constructing graph neural networks to model long-range pixel dependencies for semantic segmentation of remote sensing images in International Journal of Remote Sensing.

Visual Intelligence i Digitalisering juni 2021

June 24, 2021

Norge er verdensledende på bildeanalyse med kunstig intelligens. I juniutgaven av Digitalisering kan du lese om hvordan Visual Intelligence skal bidra til å bruke maskinlæring og kunstig intelligens for å løse viktige samfunnsutfordringer innen helse, ressursstyring og miljø.

New paper published in Machine Learning (2021)

June 24, 2021

Ahcène Boubekki, Michael Kampffmeyer, Ulf Brefeld and Robert Jenssen have published their paper Joint optimization of an autoencoder for clustering and embedding in journal Machine Learning (2021).