Dear friends,
We are excited to share with you great news: MI4People will contribute to upcoming TUM.ai Makeathon as a Challenge-Setter in MedTech track!
TUM.ai is one of Europe's largest AI-based student initiatives which aims to drive positive societal change by bringing the potential of AI to all domains. The current TUM.ai Makeathon is an international 48-hour virtual challenge with a theme AI4SocialGood in which innovators of tomorrow will develop working AI solutions for real-world problems. Students and young professionals from business, IT, medical and many other backgrounds will be working together on challenges in Education, Environment, and MedTech tracks.
The MI4People’s challenge is focused on identifying fourteen typical diseases on X-ray images of chest region. This marks the start of our second big project “General Computer Vision for Healthcare” that is focused on supporting primary care physicians in poorer developing countries by providing them with a Computer Vision System for diagnosis of diseases and injuries using various medical images, e.g., photos, X-ray images, CT and MRI scans.
Besides the actual Computer Vision models, participants are encouraged to create ideas on how to use these models in practice (e.g., app or framework). Accordingly, the Makeathon would consider special conditions in the developing world, e.g.:
Poor access to existing medical IT tools / less available computing power / restricted access to internet and cloud computing / old smartphones
Old X-ray devices, so that the quality of images might be much poorer as in the machine learning training data set
If the application/framework idea includes, e.g., using smartphone pictures of X-ray images, how different photo quality and light conditions can be considered
Challenges of integrating the system into existing processes
The focus of our challenge is not so much on the creation of a “perfect” model, but on how to apply it in practice. We welcome all students who will take up this challenge, are curious about all contributions, and are sure that they will help better health care in the developing world!
Also, Earth Day is coming up on April 22nd – this day is to celebrate our pledge to protect our planet. Accordingly, we dedicate this newsletter on topics related to our environment.
Let us together make the world a better place for all of us.
Your MI4People Team
News
Incorporating AI to Maximize Clean Energy
Renewable energy is becoming increasingly important for our planet since using fossil energy is not a sustainable option anymore due to progressing climate change. In addition, recently, the prompt adoption of renewable energy has become a geopolitical target for many countries in order to become independent of fossil energy supplies from Russia.
One of the most important renewable energy sources is wind. However, converting the wind into electricity is associated with many technical challenges. One of these challenges is so-called “wake effect” that is observable in large wind farms and means that the turbines that are located downstream cannot use the whole potential of the wind flow.
This effect can be mitigated by dynamic positioning of the upstream rotors to deflect the wake effect away from downstream turbines. However, this approach is highly complex and difficult to implement. But with help of artificial intelligence (AI) this becomes a much easier task! Vestas – one of the leaders in wind energy industry – has cooperated with Microsoft and minds.ai to create an AI model that is able to control wake effect. This model is based on the so-called reinforcement learning – a type of AI models that learn largely by trial and error, i.e., they test out different actions and get rewarded when a particular action achieve a better result.
Currently this project is still in an early stage, but it shows that intelligent algorithms are able to help the society and the industry to transform towards green energy.
Using Machine Intelligence to Spot Wildfires
One of the effects of progressing climate change is the increasing probability of wildfires. Already now, every summer the news is full of reports about wildfires all over the world. But the situation will become even worse: UN warns that wildfires will increase by a third by 2050 and by the end of the century there will likely be 50% more wildfires.
Therefore, early detection systems for wildfires have become increasingly important. Such systems can benefit from Machine Intelligence (MI). Especially Computer Vision is a very suitable technology for identification of wildfires. It can process various data, like satellite imagery, drone footage, and videos from tower-mounted cameras to observe the region of interest, spot fires, and automatically trigger alerts.
For instance, a consortium of University of Nevada, University of California, and the University of Oregon has started a program called ALERTWildfire that provides fire cameras and MI-tools to help firefighters and first responders. This system is already used in many regions across US. For more details on application of ALERTWildfire in California, also consider reading the corresponding article in Scientific America.
Tackling Micro and Nanoplastics
According to a 2017 US study, ca. 5 billion tons of plastic or ca. 60% of all plastics ever produced were discarded and are accumulating in landfills or in the natural environment. This plastic degrades and fragments into microplastics, which is then ingested by humans and animals through food and water. Some reports suggest that we consume around five grams of microplastics every week and a recent study also suggests that considerable amount of nanoplastics – further fragmentation product of microplastics – is also present in the air and we ingest this via our lungs.
While we should reduce our plastic production and consumption in the first place to fight this problem, we should also create technology to be able to filter this plastic out of the environment. Thereby, the first step is to identify micro- and nanoplastics. A team of Italian researchers has suggested a method how to use a combination of 3D coherent imaging and machine learning (ML) to achieve accurate and automatic detection of microplastics in water. Their current model is reaching 99% accuracy in filtered water samples and can identify microplastics of different size and types of plastic materials.
We hope that this and similar efforts will lead to creation of innovative solutions for filtering microplastics from environment and maybe even nanoplastics from the air.
Concluding Remarks
In this issue of our newsletter, we have seen several inspiring examples of MI applications for environmental protection. We also would like to emphasize that presented studies and prototypes are not only helping protect environment and save people's lives but can also be the foundations for sustainable and profitable business models. We hope that in the near future a lot of new innovative startups and companies will harness the power of Machine Intelligence to tackle environmental problems. If you are interested in further examples of MI applications for environmental protection, read one of our previous newsletters and our dedicated whitepaper on this topic.