Wildlife
Starting Soon

AI against Oil Spills 2

Detect small-scale oil spills on (RGB) drone video footage and estimate the volume and type of oil spilled.

ACHIEVE YOUR AI LEARNING GOALS

Premium participation: Get 1-on-1 Mentoring & Guidance

Try a free mentoring session

Applying AI to detect oil spills before they cause damage to the environment.

Despite the decreasing number of accidental oil spills in European waters, major accidental oil tanker spills still occur regularly. Large (i.e. greater than 7 tonnes) accidental oil spills account for about 10-15% of all oil that enters the ocean every year. More oil leaks into inland waterways in smaller but more frequent accidents on rivers and in ports.

Once detected, these spills are cleaned up by government agencies and specialists, using specially designed tools and vessels. Cleaning up oil spills is a costly affair, but the sooner a spill is detected, the easier it is to clean and the less it impacts the environment. Finding methods to decrease the response time to oil spills is therefore crucial to mitigating their impact.

FruitPunch AI has teamed up with Rijkswaterstaat, in charge of the management of the main infrastructure facilities in the Netherlands, and Shift Environmental to figure out how to use satellite data and drone images to detect oil spills before they get dangerous to the environment.

Info session

Our challenge partner

Detecting smaller and more frequent spills on RGB drone data

Detection of oil spills in inland ports and waterways mostly relies on visual observation. Satellite data is used to detect large-scale oil spills on open seas, but so far it is not yet possible to monitor small, freshwater spills. Building an AI model to detect these smaller, but more frequent spills as well, could potentially help notify inland spill response teams sooner and help them better safeguard water quality.

Drones offer an interesting possibility to remotely detect oil spills. Currently, they are used reactively, to quickly assess any notifications of potential spills. An expert then has to judge the images to determine the size of the spill and the type of oil. Automating the detection process would open the door to autonomous monitoring, greatly improving the scale at which environmental agencies can operate.

In AI against Oil Spills 1 we came up with more reliable ways of calculating oil spill volume so that the response team can do a closer estimation of the resources they’ll have to deploy to react efficiently.

In AI against Oil Spills 2 we will build upon this with more and better data and deploy the models in the real world doing (near-)live inference.

In this Challenge, we will

  1. Validate the model made in AI against Oil Spills 1 on new data
  2. Improve models made in AI against Oil Spills 1
  3. Build a model to estimate the oil spill volume
  4. Make sure the model can run (near-)live inference

Data we will use

  • RGB drone data of incidents in inland waterways, accurately labelled by Co-one

Results previous Challenge: AI against Oil Spills 1

In a short time, with limited resources, and scarce data, we produced results that demonstrate an AI solution to tackling inland oil spills is not a farfetched possibility. The performance of the models is indicative of the potential that AI has in tackling the problem of oil spills.

In this blog post we dive into AI against Oil Spills 1:

Blog post - AI for Oil spills

In the link below you will find all the end results of this challenge written by the Challenge participants. The folder consists of final reports, presentations, and code written for the Challenge.

End Results - AI for Oil spills

Application deadline

January 5, 2024
To application page

Timeline

FRI 5 JAN 17:00 UTC - Application Deadline

TUE 9 JAN 16:00 UTC - Challenge Kickoff

TUE 19 MAR 16:00 UTC - Challenge End

A little more