Google’s ‘AI for Social Good’ is now making an Impact in Animal Conservation!
Learn how Google is applying ML to protect a struggling, 73 member, killer whale species in the Sali
Arun C Thomas
Google’s AI for Social Good is an initiative, in which AI is used to address some of the world’s biggest challenges. The program primarily focuses on solving humanitarian and environmental challenges, using Googles AI expertise. Even though the team recognises that AI is not a silver bullet, they believes the AI can definitely help, when there is a collaborative and concentrated from all sectors of society. Predicting risk of cardiac events using retinal imagery and Quickly and accurately forecasting floods are some of the early achievements of the team.
In its recent efforts, the team is collaborating with Fisheries and Oceans Canada (DFO) to apply machine learning to protect a rare species of killer whales, also called ocras. As per the information published by the Center for Whale Research, only 73 members of this whale species are left and they are struggling to survive. Disturbance caused by human activities, passing vessels, contaminants and scarcity of prey are the major contributors of threat in the Salish Sea, where the Southern Resident Killer Whales are found, as per DFO.
Deep Neural Networks, trained with DFO provided 1,800 hours of underwater audio and 68,000 labels of the origin of ocras, is used to analyse the live sounds that DFO monitors. These trained neural networks are used to identify sound of ocras from the live sound feeds, which is obtained from 12 locations within the Southern Resident Killer Whales’ habitat. Once the sound of the the killer whale is recognised, the live location from the sound feed is mapped on to the Rainforest Connection web interface. This allows live tracking of whales.
The availability of high quality labelled data set is a key factor in contributing to the success of these kinds of projects, as the the data-set (underwater audio) itself can be rare to find. A high quality labelled data set means that, the computational resource spend on training Deep Neural Network models is spent effectively, as the time taken for training can be really high for these kinds of application where the input data set is relatively complex.
Along with the quality of data-sets, the amount of well labelled data-sets used to train the model also can contribute to the success as the model will be able to do inference precisely on a wide range of data sets. Also in this case, inference is done form input stream from multiple locations which maybe computational complex, but will provide exact location of whales in approximate real time.
Additionally the data obtained is used for further analysis, to obtain interesting patterns which can be helpful, not just for monitoring, but also to treat whales that are injured, sick or distressed. The system is also equipped for special scenarios like, in the case of an oil spill, will be useful to authorities to identify the position of the whales and help to redirect them, using specialised tools, to a location of safety. Even live alerts are provided to key stake holders using an app Rainforest Connection developed. All of these are contributing towards improved conservation of this endangered whale species.
This collaborative effort also shows how Artificial Intelligence can be useful in different fields of Science and Research, like here in the case of Bioacoustics.
Bioacoustics is a cross-disciplinary science that combines biology and acoustics. Usually it refers to the investigation of sound production, dispersion and reception in animals (including humans).
While this is a great step towards animal conservation, this also proves that how advancements in Artificial Intelligence and Machine Learning will be useful for all the species in this universe. As the team at Google states,
We hope that advances in bioacoustics technology using AI can make a difference in animal conservation.
I believe that these are just baby steps that we are taking, while imagining the actual capability of AI/ML based technologies. These small steps count as we need to fight back against incidents like Australia’s Wildfire,s where an estimated one billion animals had died. Use of advances technologies and action at correct time can make a lot of difference. After all a life is a life, it is beautiful, and it needs to be preserved and its our responsibility.
Thanks to each and everyone at AI for Social Good at Google for your efforts and good luck for the future.
This article was published by Arun C thomas on medium.
Arun C Thomas