About tinyML for Good
Technological advancements are transforming the way we live, work, and connect with the world around us. Tiny machine learning and artificial intelligence are enabling on-device sensor data analytics at extremely low power and with privacy built in by design, already showing great potential to make positive contributions to the United Nations Sustainable Development Goals, particularly in low-resource settings. In this tinyML for Good conference, we will be showcasing inspirational thinkers from the education, healthcare, environment and development sector*, as well as real-life, practical examples of tinyML in the world. Come along and learn what tinyML is and how it can be used in your programme, organization or region. No technical experience required.
- Keynote: Kate Kallot, Nvidia 7:00 am – 7:45 am Pacific Time
- Lightning talks (5 mins each) in the three focus areas STEM, Healthcare, Sustainability/Climate 7:45 am – 9:50 am
- Call to Action Wrap-up 9:50 am – 10:00 am
- “Q&A overflow and networking”: 10:00 am – 10:30 am
7:00 am to 7:45 am
Introduction and Workshop Objectives, Evgeni Gousev, Qualcomm
Keynote: Kate Kallot, Nvidia
7:45 am to 8:40 am
Lightning talks - STEM
Each talk is 5 minutes including questions.
Session Moderator: Alessandro GRANDE, Head of Product, Edge Impulse
Session Moderator: Vijay JANAPA REDDI, Associate Professor, Harvard University
Session Moderator: Brian PLANCHER, Ph.D. Candidate, Harvard John A. Paulson School of Engineering and Applied Sciences
How tinyML can redefine Computing Education
Robert LEEMAN, Education Solutions Manager, Arm
Computing Education has an engagement, image and diversity problem. The highly practical and creative nature of the subject is not adequately realized in formal curricula and risks disengaging learners, especially those from already unrepresented minorities. Practical/Physical Computing, using tinyML is a potential solution. This blend of pedagogy, practice, resources, technology and training offers a way forward in re-engaging learners into STEM subjects by redefining what Project Based Learning and Physical/Practical Computing means to both educators and learners by applying Computational Techniques using modern physical devices to gather and interrogate data in an artifact that solves a real, engaging and relevant problem. This approach makes Computing and Data Science come alive but requires a perspective shift and often training for educators to fully realize the potential of this teaching approach.
Empowering students around the world to tackle the Global Goals with the BBC micro:bit
Jonny AUSTIN, Chief Technology Officer, Micro:bit Educational Foundation
There are now more than 5 million micro:bits in the hands of students around the world. We’ll look at how designing a device to be easy to use for education, and focusing students on the big challenges facing the world helps create the innovators of the future.
tinyMLedu: Widening Access to Applied Machine Learning
Vijay JANAPA REDDI, Associate Professor, Harvard University
Our mission is to widen access to applied machine learning technologies by establishing best practices in education, research and outreach. We are focused on building an international coalition of students, teachers, and researchers focused on advancing embedded machine learning, developing and sharing high-quality, open-access educational materials globally. To date, we have successfully created the Tiny Machine Learning course series on edX, a massively open online course with over 50,000 students that have enrolled in the program, many for free. In addition, we helped launch two courses derived from the HarvardX materials, taught in Portuguese in Brazil. We have also held an outreach workshop for high school teachers and students of the Navajo nation and launched an Academic Network of over 20 universities from around the globe. Moving forward, we want to grow our impact by developing more workshops and teaching courses in more languages, targeting an even broader audience, to introduce the world to tinyML.
TinyML4D: improving access to tinyML through a global academic network
Marco ZENNARO, Research Scientist, the Abdus Salam International Centre for Theoretical Physics
Machine Learning (ML) has a huge potential to tackle societal issues in diverse fields including agriculture, conservation, and healthcare. TinyML is a cutting-edge field that brings the transformative power of machine learning (ML) to small low-power and low-cost computing devices. The advent of TinyML offers new opportunities for complex on-device ML applications and research. For example, as these devices are inexpensive, students can use them for their school practicals and experiments, and with their low power requirements they can be widely deployed to positively impact society. TheTinyML4D working group is building a network of academic institutions, based in Developing Countries, interested in expanding access to Applied Machine Learning by establishing best practices in education. We aim to ultimately develop a community of researchers and practitioners focused on both improving access to TinyML education and enabling innovative solutions for the unique challenges faced by Developing Countries.
tinyML4K12: creating a tinyML education ecosystem for young learners
Hal SPEED, Head of North America, Robotical
Expanding tinyML education into primary and secondary schools (K-12) requires the development of an end-to-end pipeline that is appropriate for school-aged children. The tinyML4K12 working group is collaborating with education and industry partners to combine computer science education software and the physical computing ecosystem to enable an easy learning experience for creating, deploying, and using tinyML models. This pipeline will enable the creation of additional materials that can be used across the globe for students of all ages.
tinyML4STEM: using tinyML for Neuroscience in K12
Greg GAGE, CEO , Backyard Brains
Combining tinyML and Neuroscience enables exciting, hands-on STEM education experiences for the K12 audience. In this talk we will describe our successful effort piloting this exciting collaboration through Backyard Brains this past summer.
tinyMLedu Outreach: Embedded Machine Learning for the Navajo Nation
Brian PLANCHER, Ph.D. Candidate, Harvard John A. Paulson School of Engineering and Applied Sciences
tinyMLedu’s Outreach efforts focus on workshops that encourage broader participation in embedded machine learning. As a part of these efforts, and in collaboration with Navajo Technical University, the Harvard John A. Paulson School of Engineering and Applied Sciences, Google, and Edge Impulse, tinyMLedu ran a 4-day, hands-on workshop for 50 high school teachers and students, affiliated with the Navajo nation, exploring real-world applications of artificial intelligence through hands-on examples of tiny Machine Learning. By the end of the workshop we had not only surveyed the fields and a series of applications of artificial intelligence and (tiny) machine learning, but had also worked through the entire applied machine learning flow from collected data, to design and training models, to converting and deploying them onto edge devices.
TinyML for AI Robotics Applications in Education
Vincent KOK, Product & Sales Manager, UBTECH Robotics
With the rapid evolution of technology and the requirements of SDG #4 (Quality Education), STEAM educators must equip themselves with modern skills and tools sets. Tools such as Seeed Studio’s Codecraft graphical block-based programming platform powered by Edge Impulse enabled educators to easily create TinyML projects on Wio Terminal hardware. Teachers can then share their acquired knowledge with their students in making TinyML related projects to solve real-life problems. Beyond graphical block-based programming, educators could also leverage the model generated and use it in a text-based coding environment. This allows the scalability of the TinyML projects in various scenarios. For this talk, Vincent will also share his Hackster.io’s featured TinyML robotics project (“Voice Activated Robot Car on Microcontroller with TinyML”) as an example of how STEAM educators can quickly prototype TinyML projects.
Smart Birdfeeder – A TinyML experience with young kids
Parker ZHANG, MTS Staff Senior, Qualcomm Canada
9 young kids, aging from 5 to 9, out of 5 families, spent one week together in this summer working on this Smart Birdfeeder project. By designing the feeder from scratch, building up using Lego piece by piece, learning fundamentals, collecting training data, assembling the TinyML components, and eventually they built their own birdfeeders which can detect “greedy squirrels” and scare them away. To our great surprise, the project won the “Honorable Mention” award from Eyes on Edge: tinyML Vision Challenge, which is the best summer gift for the kids and families. This exciting journey demonstrated that TinyML is indeed an accessible and fun way for kids to experience AI in their early age.
8:40 am to 9:10 am
Lightning talks - Healthcare
Each talk is 5 minutes including questions.
Session Moderator: Steve WHALLEY, CEO, Strategic World Ventures
Pneumonia Detection using Embedded Machine Learning
Arijit DAS, Ambassador, Edge Impulse
Pneumonia has troubled the world for decades, with deaths mounting when early detection is missed. The WHO estimates that each year the number of people affected by pneumonia is nearly 450 million, with the number of deaths reaching 4 million per year, mostly in the developing world. These numbers represent a lot of pain, lost dreams and grief for people everywhere, and can be mitigated. So far, the detection of pneumonia is achieved by using x-rays, chest scans or extracting blood serum in extreme cases. BUT this requires time, resources, skilled personnel, and it is expensive! I’ve been working on a solution that harnesses low-cost hardware and software that can deliver effective detection for people everywhere, regardless of location, skill and affordability.
Towards TinyML Solutions for Extreme Heat Sensing for Urban Climate Science
Suren JAYASURIYA, Assistant Professor, Arizona State University
Extreme heat puts tremendous stress on human health and limits people’s ability to work, travel, and socialize outdoors. To strategically deploy heat mitigation strategies in cities, thermal conditions must be assessed in the context of human exposure and space use. We introduce MaRTiny, a novel, low-cost thermal and visual sensing device that detects how people use outdoor spaces when it is hot. The prototype collects meteorological data, concurrently counts the number of people in the shade and sun, and streams the results to an AWS server. MaRTiny lays the foundation for fundamental research in urban climate science to investigate how people use public spaces under extreme heat. This research will inform active shade management and climate action planning in warming cities, and this talk will serve as a call to action for the TinyML community to help stakeholders with these goals.
How Dr Car will save healthcare personnel from coronavirus
Mouhamadou Lamine KEBE, System, network and telecommunication engineer, Ecole Supérieure Polytechnique de Dakar
Moussa SECK , Engineer in networks and telecommunications, Ecole Supérieure Polytechnique de Dakar
During covid19 , many African countries were facing a lack of equipment, hospitals were overloaded and medical staff were very exposed to the virus. Dr CAR is a medical device that embeds smart sensors to do telemedicine and remote care. we use edge solutions to do local data processing and accelerate decision making.
Offline Prediction of Cholera in Rural Communal Tap Waters Using Edge AI inference
Marvin OGORE, Student, The University of Rwanda
Africa accounts for 54% of the world disease burden due to the lack of access to safe drinking water, with the majority of rural area populations or endemic zones getting access to water through potentially unsafe communal water taps. Unfortunately, the expensive laboratory processes and resources used in water processing centers to detect water-borne diseases like cholera cannot be massively deployed on all those taps to guarantee safe water for everyone, anywhere at any time. Thanks to the integration of Internet of Things (IoT) and Artificial Intelligence (AI), the prediction of water-bone cholera can be done by monitoring water’s physicochemical patterns. However, related state of the art IoT/AI solutions rely on a cloud-centric architecture with edge water parameter sensors sending collected data to the cloud for inference. Unfortunately, anytime wireless connectivity is not always guaranteed in rural areas, but also it is very consuming in terms of energy for a system expected to run several years without maintenance. Last but not least, low latency detection is mandatory to warn the tap user on time. Our work presents a prototyping design and development of an offline edge AI rapid water-bone cholera detector kit pluggable into existing taps to lower the cost of mass deployment. Our simulation results and onboard results in an embedded context show a good accuracy of edge inference with respect to live cloud classification.
9:10 am to 9:50 am
Lightning talks - Sustainability
Each talk is 5 minutes including questions.
Session Moderator: Christopher B. ROGERS, CEO, SensiML Corp
Ribbit Network: The world’s largest crowdsourced network of open-source, low-cost, CO2 Gas Detection Sensors
Keenan JOHNSON, Founder, Ribbit Network
We know that increased levels of atmospheric gasses like CO2 are the primary causes of climate change and humanity is pumping them into the atmosphere at an unprecedented rate! It would be reasonable to assume that scientists have a map of the world that can tell us exactly what the emissions levels are at any spot on the planet. Unfortunately, that map doesn’t exist. The Ribbit Network was created to generate this missing map of greenhouse gas emissions. By creating and deploying the world’s largest, grassroots network of CO2 sensors, the network empowers anyone to join in the work on climate and provide data for informed climate action.
Groundwater Wells: The Untapped Edge Use Case
Doug STANDLEY, Founder and CEO, Niolabs
This talk, and the TinyML community call to action, highlights the lessons learned digitizing a farm with edge computing over the past 5-years, the benchmark results from the Gen.1 implementation, and the exciting expectations and opportunities from the in-process Gen. 2 re-imagination-architecture-implementation that may include members of the TinyML community.
Acoustic Monitoring of Ecosystems
Ciira MAINA, Senior Lecturer, Dedan Kimathi University of Technology
Emasi COLLINS, Undergraduate research assistant at the Centre for Data Science and Artificial Intelligence, Dedan Kimathi of Technology
Over the past century, the world has experienced an increase in human activities and climate change
that have resulted in severe degradation of biodiversity. Swift actions need to be taken in order to conserve our
ecosystems. Development of single board computers and microcontrollers enable us to collect data from the
ecosystems remotely. This makes it easier to monitor wildlife without the need to carry out the inefficient
physical surveys. In this presentation, we will demonstrate how we can use the Raspberry Pi Pico to perform
bird audio detection which can be used to log acoustic events of birds with the aim to monitor the ecosystems.
How tinyML powered precision beekeeping could help save the bees and improve honeybee yields
Clinton ODUOR, Co-founder, Rhions Lab
Jackline TUM, Ambassador, She Code Africa
Every third bite of food we take depends on bees’ pollination, yet we lose an unprecedented 30% of our bee colonies every year. In countries across Africa, honeybees could help protect wildlife, grow food crops, and make money. However, sustaining apiaries in the face of parasite infection, rodent invasion, and hive heists has become a hinderance to apiary profitability.
Ibees is a solution to help save the bees, and help beekeepers monitor the health of their beehives in real-time on their phones. It incorporates arm-based tinyML powered devices mounted on beehives that collects temperature, humidity, acoustic and inertial sensor data. The ML model running on the devices processes the data and sends a message to farmers’ mobile phones when an anomaly is detected.
The massive amount of data collected from the network of the connected intelligent beehives could be used by researchers, conservationists, and governments for analytics to find insights on the possible solutions to the rapid bee colony decline.
MULTI-SENSOR FISHERY DEVICE
Barke ABDALLAH UKUSI, Developer, The State University of Zanzibar
Multi-sensor fishery device is designed to use artificial intelligence technology (AI) that is intended to offer solutions to very long-time unsolved problems in fisheries sector of Zanzibar. The device will be loaded with tiny sensors that will detect and provide information on where about to fish, average size of fish (within the acceptable size) to be stripped, also safety and rescue information to fishermen and the disaster management unit respectively. Further, the device will be a fish aggregate that luring more fishes to be extracted. The multi-sensor fishery device will be linked to server unit (Internet of thing – IoT) through LORA for storing information necessary to inform the government on fisheries decision-making and policy formulation, for better and sustainable management of the fish resources in Zanzibar. Fisheries information is lacking in Zanzibar and collecting such information is laborious and expensive, thus multi-sensor fishery device will be the solution.
Kids can save animals
Kate Gilman WILLIAMS, CEO & Founder, Kids Can Save Animals
One aspect of Kids Can Save Animals is her podcast,Club 15,named so because we lose one elephant from our planet every 15 minutes. Club 15 was created in partnership with Sarah Maston and Daisuke Nakahara of Microsoft, and is an extension of Project 15 from Microsoft. Each episode of Club 15 connects youth with top scientists, technologists and conservationists working on the ground to save animals from extinction. Microsoft contributed a learning lab for Club 15 so kids can learn how Computer Vision and Artificial Intelligence (AI) are collecting data and how technology is on the forefront of saving animals.
9:50 am to 10:00 am
Call to Action Wrap-up
10:00 am to 10:30 am
"Q&A overflow and networking"
Session Moderator: Alessandro GRANDE, Head of Product, Edge Impulse
Schedule subject to change without notice.
Qualcomm Research, USA
Christopher B. ROGERS
Strategic World Ventures
Micro:bit Educational Foundation
Vijay JANAPA REDDI
the Abdus Salam International Centre for Theoretical Physics
Kid Spark Education
Harvard John A. Paulson School of Engineering and Applied Sciences
Arizona State University
Mouhamadou Lamine KEBE
Ecole Supérieure Polytechnique de Dakar
Ecole Supérieure Polytechnique de Dakar
The University of Rwanda
Dedan Kimathi University of Technology
Dedan Kimathi of Technology
She Code Africa
Barke ABDALLAH UKUSI
The State University of Zanzibar
Kate Gilman WILLIAMS
Kids Can Save Animals