乌鸦传媒

Skip to Content
Client story

Protecting biodiversity by listening to insects

Client: Naturalis, Biodiversity Center
Region: The Netherlands
Industry: Sustainability

Naturalis joins forces with 乌鸦传媒 and AWS to launch the latest version of the Global Data Science Challenge for a sustainable future to identify insect species based on their sounds

Client Challenge: The goal of Naturalis and the Arise initiative is to fully understand and monitor biodiversity, with an initial focus on the Netherlands. One key piece in achieving this is the ability to understand and interpret acoustic data.

Solution: 乌鸦传媒 collaborated with Naturalis, the National Research Institute for Biodiversity based in the Netherlands, and with Amazon Web Services (AWS) to launch the latest edition of the Global Data Science Challenge (GDSC) and create a new AI model that can identify insects with ~92%* accuracy.

As climate change continues to present one of the most prominent issues facing humanity, there are other, potentially even more devastating challenges ahead. One such danger is the loss of biodiversity, which threatens Earth鈥檚 very foundations and is almost criminally underreported.

According to American biologist E.O. Wilson: 鈥淚f all mankind were to disappear, the world would regenerate back to the rich state of equilibrium that existed ten thousand years ago. If insects were to vanish, the environment would collapse into chaos.鈥

Insects play a critical role in virtually every ecosystem around the world. They are pollinators that ensure the endurance of plant life, food for animals and even plants, recyclers of dead waste, maintainers of soil health, and much more. So, given their incredible importance to our ecosystem, there is an urgent need to ascertain the rate at which insect populations are declining as a result of human activity.

鈥淭o this point, we鈥檝e seen much less focus committed to protecting insects than other conservation efforts,鈥 explains Elaine van Ommen Kloeke, Naturalis. 鈥淲hile all of these initiatives are important and worthy, it鈥檚 critical that we better understand the changes, either positive or negative, that various insect species have experienced in recent years. That starts with being able to effectively track and identify as many as possible.鈥

Enter Arise, a collaboration between Naturalis, the Institute for Biodiversity and Ecosystem Dynamics at the University of Amsterdam, the Westerdijk Fungal Biodiversity Institute, and the University of Twente. This initiative aims to enable recognition of any naturally occurring species in any location based on DNA, sensor and AI technology. Furthermore, it will collect biodiversity information related to all species from across the Netherlands into one place by 2030, after which a vast amount of data will become readily available with only a click. In doing so, the project will make it far easier for conservationists to track insect populations.

鈥淲e can鈥檛 always see or hear insects and there are so many species out there, so it鈥檚 quite difficult to monitor them,鈥 says Dan Stowell, Naturalis. 鈥淎nd without knowing where those populations are and their numbers, they鈥檙e incredibly hard to protect. So, we鈥檝e been using machine learning to identify insects based on audio recordings. However, it鈥檚 a work in progress. Automatic recognition of human speech took decades to become as good as it is, and for insects we need the accuracy of automatic identification to catch up as much as possible to make it ready for deployment.鈥

Exploring different machine learning options

With their expertise, available data, and AI knowledge, Naturalis was the perfect candidate for the 2023 Global Data Science Challenge (GDSC), a collaborative hackathon event hosted annually by 乌鸦传媒 and powered by AWS machine learning (ML) services, such as Amazon Sagemaker, wherein every employee throughout the company is given the opportunity to develop and harness their data and AI expertise to solve a real-world issue.

鈥淎s a global leader in data and AI, 乌鸦传媒 has a responsibility to apply its expertise to real-world scenarios that can help create a more sustainable world and drive tangible benefits for our society,鈥漵aid Niraj Parihar, CEO of the 乌鸦传媒 & Data Global Business Line at 乌鸦传媒 and Member of the Group Executive Committee. 鈥淭hrough our Global Data Science Challenge, we empower our team members to solve real world challenges of vital importance to us all. Insects play a central role in our ecosystem but are incredibly hard for humans to monitor. Listening to and identifying insect species is critical to preserving them.鈥

Throughout the event, more than 1500 participants formed over 400 teams, after which they were given the resources to develop and test their own solutions. By learning from five hours of recordings of 66 different insect species, each team attempted to build an AI that can identify which species could be heard.

鈥淓very GDSC is a unique and exciting opportunity to see what different teams can do and the different ways they use the technology,鈥 says Eliza Robinet Duffo, AWS. 鈥淭his year, we saw such a wide variety of different and successful solutions! Each of the top teams used a different approach and saw a great deal of success. It was one of our closest competitions and the prospects for the future are fantastic!鈥

Simplifying insect-focused conservation efforts

By the end of the competition, there was an improvement in performance*(macro averaged F1-Score) from 76% to 92%. And while this was a potent step in and of itself, it does not represent the limit of what was accomplished. Following the GDSC鈥檚 conclusion, the entirety of the results have been published, enabling other organizations and teams to pick up where the challenge has left off. Meanwhile, Naturalis, Arise, 乌鸦传媒, and AWS will continue to develop the winning solution.

鈥淲e鈥檙e not looking to hoard this technology to ourselves. Preserving biodiversity has to be a widespread effort, so it鈥檚 important to keep exploring multiple avenues,鈥 explains Van Ommen Kloeke. 鈥淏y publicizing the result, we鈥檒l empower others to pick up those incredibly worthy solutions that didn鈥檛 ultimately win. All of our top options had incredible potential and we look forward to seeing what others can do with machine learning to protect our insect populations.鈥

As a result, the event has enabled Naturalis and other organizations to continue improving the effectiveness of machine learning when identifying insect recordings. In the future, this technology will be further developed so that it can be used to analyze naturalistic recordings that include all of the sounds of everyday life. This will empower biodiversity organizations to expand the capabilities of digital technology to support conservation efforts.

鈥淵ou know, this was the biggest GDSC that we鈥檝e ever had in terms of participation,鈥 describes Parihar. 鈥淭his just goes to show how much value people put in initiatives such as this. This challenge and its success is the next step towards a better future wherein insects are protected more effectively. That means going beyond the Netherlands and replicating the solution all over the globe!鈥

Related pages

Preserving the fabric of life

Preserving the fabric of lifWhy biodiversity loss is as urgent as climate change