I’m putting together a SigInt hardware library (blueprints, 3d printables, BoM, code) for radio hacking for all to download, make, and use!
I would like to create a set of devices using inexpensive methodologies (COTS hardware, modules, 3d printing, Linux, open source software stacks, some electronic boards designed as needed) that enable fellow hackers to be able to investigate and interrogate radio-based hardware of all sorts.
1. Identify devices that is either prohibitively expensive or not created (examples of devices would be: radio-direction finding hand-held device, NLJD, updated SigInt tablet, radio mic/camera detection and location system)
2. Create proof-of-concept that does the thing identified
3. Provide full bill of materials, 3d printables, build steps, code, and notes to reproduce fully
4. Build a small amount of them (depending on cost, naturally), and get them in the hands of the community at no cost.
This project will provide new or remixed ideas on radio SDR operations, but ideas are the small thing.
This project will also provide full reproducible build instructions for said devices. I’ve already done this with the SigInt Tablet, and the autotuning antenna array.
Lastly, I would like to target people in the community (or perhaps semi-random) for people to receive the platform. The quantity made for this would highly depend on cost of hardware.
The FCC has announced a band of spectrum known as CBRS (Citizens Broadband Radio Service), which offers spectrum access across three tiers. The last and lonely tier is tier 3: General Authorized Access (GAA). [ https://www.fcc.gov/35-ghz-band-overview ]
The intent of this project is to explore community driven mobile communications in the 5G space making use of this spectrum, which requires no licensing or purchasing.
By building a community 5G network, participants (neighbors, kids, students, whoever wants to help) will learn how 5G networks work, how to make them more secure, and gain valuable skills that will have immediate use.
I hope to learn about and teach others many important skills:
1. Demystify 5G networks by learning by example
2. Learn how to deploy and manage 5G networks
3. Provide a free, or minimal cost, fixed wireless network for my community
4. Contribute to many open source programs, including Open5GS and Magma
5. Hack and study the cyber security risks of 5G networks.
6. Take advantage of the CBRS spectrum. As citizens if we don’t use it, it will get taken by service providers and others instead.
We can start building a community/municpal internet service provider using fixed 5G solutions!
We provide summer, after-school and weekend STEM classes, events, workshops, training and competitions that cater to Youth (ages 6-17) in the Washington DC metropolitan area in after-school programs in nontraditional venues. We are looking to support 75 students in the robotics, coding and competition with VEX robotics training and education for a year long program November to March 2022. We have past performance for 10 years in the field of supporting youth
Project goals: Expose up to 50 students K-12 in a community base robotics/coding (languages – Block-based, Python, and C++.) and competition for 9 weeks.
Student assessment before, during and after the program will measure STEM knowledge, robotics equipment and programming skills.
Family engagement Survey what parents involvement to
a. Improve program responsiveness to children and their families
b. Improve the design and delivery of programs and services.
c. Strengthen the connection between families and communities and the programs that serve them
END Result:( 4 major goals )of project is the following
1. Achieve basic knowledge and skills in the areas of STEM, engineering, design, programming, and manufacturing/assembly.
2. Understand basic programming skills by training students in Robot C.
3. Increase problem solving skills and an appreciation of creativity, invention, and
innovation in the design process through the use of CAD assisted robot design and
4. Understand of that the work of engineers and scientists is rewarding and that these
fields can be accessible for students of all demographics and abilities.
5. Foster an appreciation of teamwork, leadership, sportsmanship and mentorship.
Expose underrepresented k-12 students from various public, community homeschools and private education in DC MD and VA programs and practical hands-on STEM classes and competitions.
Globally, around 59% of the world’s waste ends up in a landfill, which leads to further environmental complications with pollutants leaching out into runoff or groundwater, land use change, and greenhouse gas emissions. Furthermore, over 8 million tons of plastic waste enters the oceans annually and only 20 rivers in developing countries account for 67% of the outflows due to the lack of centralized waste management infrastructure. Nearly 80% of plastics entering our ecosystems comes from land based sources. This waste can be used as resources for empowering people and communities instead of polluting ecosystems. Since the creation of centralized waste infrastructure is expensive, a different decentralized approach could empower people in places which lack proper management.
At Mycelium, we believe this problem can be addressed by implementing modern machine learning technology into standard waste receptacles, so that waste can be identified and sorted upon entering the receptacle. Instead of attempting to change human and societal behavior, we aim to decentralize waste management to make waste sorting convenient. Therein lies our goal for this project, and furthermore we aim to share our machine learning trash image dataset and trained model with the world via an open-source license. The ultimate goal is to create a compact automated trash collection system for separating waste by type for accelerating the circular economy. Such an automated system would enable efficient recovery of sorted waste for recycling or repurposing goals. However, this grant will exclusively be used to create an open-source dataset with a fully trained model which anyone around the globe can use to implement their own automated waste sorting systems.
Project Goals: Our main goal is to create an open-source dataset of images of trash of every type with a trained neural network for ultimately creating an intelligent trash can which can automatically separate waste. This grant will only be used for creating the dataset along with a custom-made trash can and a machine learning model for categorical identification of waste.
Measurement: Each waste category will have at least 2000 images taken with an installed camera inside our custom trash can. This will provide sufficient images to train a robust convolutional neural network model with a minimum threshold accuracy of 98 percent with high precision and recall scores. The deep learning model will be trained using the TensorFlow framework in Python and will be licensed as open-source on GitHub.
End Results: Our end results will be an extensive open source dataset containing all categories of domestic waste (plastics, cardboard, paper, styrofoam, aluminum, food, etc) and a trained model, shared on GitHub that anyone can use to add “intelligence” to their own trash can.
The project’s end results of a machine learning ready dataset and a trained model for categorical waste identification can benefit recycling companies as the sorted waste can be a direct input to their factories. Furthermore, this project will have a far reach impact since our dataset and model will be open-source so anybody around the globe can use our work without any cost. Mycelium’s vision is to enable and accelerate the circular economy with open-source technologies. We strongly believe that this project can benefit cities, especially in developing countries, to sort and use their waste instead of letting it flow into and damaging our ecosystems.
Ultimately, this project will lead to the creation of an automated trash can which will sort waste by its category. This will initially benefit the local community of Huntsville, Alabama, by reducing the amount of waste entering the local incinerator and landfill, and salvaging that which can be recycled. Local recycling facilities could additionally benefit by reducing the amount of hand sorting materials.
ScaleDown is an open-source framework and community that provides free training and mentorship as well as fosters collaboration and innovation in TinyML.
TinyML devices like Arduinos are resource-constrained. This means that they are usually small, battery powered and have low computation power and memory. Deploying neural networks on such devices is difficult due to how large they are in terms of their memory footprint and the number of operations needed to execute them. So we need to optimize and scale down neural networks to deploy them. This machine learning paradigm is known as TinyML. It is estimated that there will be more than 60 billion IoT devices worldwide by 2035. Many of them will be equipped with TinyML capabilities.
However there is a wide gap in the community for learning resources and tools to help beginners and early career professionals learn about TinyML, experiment and make products to build a portfolio to get jobs. To help people better understand this field, we do community work like hosting workshops, study groups and talks. We also create free learning resources like books and courses.
There are many TinyML optimization algorithms, but the tools landscape is fragmented with each framework supporting different algorithms and only on specific model types. Moreover, they only support a few algorithms and not the latest, better performing algorithms. This makes these tools less accessible to disadvantaged groups who may not have access to those software or hardware. Moreover, they are not easy for beginners to learn. Scaledown is bridging that gap by building a framework that can take models trained in any framework, optimize it using the latest algorithms, and then deploy to TinyML devices. Finally, we help grow the field by doing research and giving opportunities for collaboration within the community. Link: https://scaledown-team.github.io/
Our mission at ScaleDown is to educate students and early career professionals about TinyML and to build tools to make developing TinyML applications easier (Links to all our resources are given in the Community Benefit answer). Our goals for this project and how we will measure them are:
1. Teach more people TinyML: TinyML combines techniques from machine learning, embedded systems and optimization to build applications. This means that someone who works in this field needs to have a wide breadth and depth of knowledge in each of these fields. This is often the biggest barrier to entry for people trying to learn TinyML, since most people usually have knowledge in one of those areas. Our primary goal is to educate more people about TinyML by teaching these topics in a simple and lucid way so as to reduce the barriers to entry and make this area more accessible. To measure this goal, we will count the number of people we train in this field. Our previous study group reached more than 300 people. We will conduct another study group and a few more workshops (2 already scheduled) to try to help a total of at least 1000 people by the end of this year. Further, we wish to open-source more TinyML learning resources like books and blogs. We currently try to write at least 1 blog per week and are working on 2 books that we want to finish by the end of this year.
2. Create a support network for TinyML: Since TinyML requires knowledge in a lot of fields, when you are getting started, it is easy to get stuck on a problem and not be able to solve it (especially if the problems are hardware related). Further, you may also want to collaborate with other people in the field in projects, or get feedback on projects you are working on. By building a TinyML community, we can create a support network to help beginners and early career professionals in this field.
To measure this goal, we will count the number of active members we have as a part of our meetup and slack groups. <todo add goals>
3. Increase career or internship opportunities in TinyML: Another advantage of building a community is that they can support each other and give advice and resources when searching for career opportunities. We hope to help at least 2 people get a job or internship in TinyML by the end of this year.
4. Build simpler tools for working on TinyML: At the moment, tools for building TinyML applications are not helpful for beginners to get into the field. The tools landscape is fragmented with different frameworks supporting different algorithms and only models trained with their framework. Moreover, they only support a few algorithms and not the latest, better performing algorithms. At ScaleDown, we are attempting to bridge that gap and build a framework that helps you take models trained in any framework, optimize it using the latest algorithms, and then deploy it to TinyML devices. We are designing the framework to be simple enough for beginners to learn no matter what hardware they have or framework they are familiar with. At the moment we support 3 different frameworks, 2 types of hardwares and 5 types of algorithms. Our goal is that by the end of this year we want to add support for 10 more algorithms, 10 more hardwares, release a v1 of our framework and get at least 1000 downloads. We will measure these numbers to see if we have hit our goal.
5. Increase collaboration and innovation in TinyML research: Most of the latest TinyML work is released in the form of research papers. These can be very dense, convoluted and hard to understand for beginners making it even more difficult for beginners to keep up with the latest in the field. As a part of our initiatives, we do regular paper readings and publish paper reviews to make these more accessible to beginners. However, we also want our community to further the field by doing research and publishing papers themselves. We taught paper reading and how to write papers in our study groups to help them get started in this.
Our goal is to publish at least 2 research papers in this field by the end of the year. These two papers are already works in progress.
6. Increase accessibility to TinyML tools and products: A major component of TinyML is access to hardware. While most TinyML hardware are microcontrollers and hence relatively cheap, purchasing them can still be a huge burden especially for people from disadvantaged communities. To help with this, we wish to start a “Hardware Library” from where people can borrow hardware that they want to work with and then return them for a different hardware when they are done. This will help people build a portfolio of projects with multiple hardwares and tools that they can then leverage to get career or internship opportunities. We will start this library in Singapore first and then expand to other countries.To meet these goals, we will do the following:
1. Conduct more workshops and study groups. All of these will be posted to YouTube for people to refer to later as well.
2. Collaborate with other communities like Women Who Code, PyLadies, PyCon and Google Developer Groups
3. Publish TinyML educational content
4. Build a network of TinyML professionals
5. Build the ScaleDown framework for wider adoption of TinyML by reducing the barrier to entry for beginners into this field.
6. Increase research work in the community
7. Build projects
TinyML is unlike other Machine Learning communities. TinyML enthusiasts need to learn not only about machine learning, but also electronics and embedded systems so that they can optimise it for microcontrollers. This means that for someone to complete a TinyML project, they will need to have access to software to optimize their models and to hardware devices to test their models on.
The main way we will benefit the community is by conducting workshops and study groups to teach people about TinyML and how to build TinyML projects. Secondly, to reduce the barrier to entry for beginners to TinyML, we are creating a framework that people can use to optimize and deploy models. Finally, we want to start a hardware library through which more people in the community can get access to hardware to test their projects on.
In this manner, we are moving towards democratising TinyML Education all across the world.
Here are some examples of community work that we have already done:
A talk on how to deploy your models to Intel devices: https://drive.google.com/file/d/1c8q7f5uEMZOp12Up-2weHflT0KB741Iw/view?usp=sharingA
Study group content (includes YouTube recordings and Slides): https://github.com/scaledown-team/study-group
TinyML Blogs: https://scaledown-team.github.io/blog/
TinyML Books and Courses: https://scaledown-team.github.io/practical_tinyml_book/intro.html
ScaleDown framework: https://scaledown-team.github.io/
Across the board, scientists, researchers, and hackers alike have found themselves tailoring existing equipment or scratch building CNC machines to explore custom automation applications. In the last decade, we’ve seen a hoard of original machine modifications and designs that harness cnc precision to deposit cellular media, mill circuit boards, and extrude pancake batter. But while the excitement to explore new workflows with machines is present, the bar for designing and building them is high–too high for the scientist or hacker looking to focus on their application.
Jubilee is a desktop open source toolchanging CNC machine design that makes customization a first class feature. With a proven mechanical design, building a Jubilee enables machine builders to focus on the tool and software elements of their project instead of belaboring the nuances of the motion. Unlike most open source projects, however, Jubilee was designed to be self-produced anywhere in the world by builders of all skill levels. Its documentation features a rich collection of colored step by step instructions, and its Bill of Materials includes several alternates for an international audience of builders.
In the last two years, Jubilee has spawned over 150 builds and over 1700 people around the world who congregate on the project’s community Discord server. Here, Jubilee builders share tools, insights, and even ship each other parts in a friendly environment of shared project interest. These builders have extended Jubilee to applications such as pen plotting, pipetting, multi-material 3d printing, and milling, they continue to drive the project forward to a number of original applications.
Finally, to date, Jubilee is not a commercial product. While some components are available by some community vendors, it largely stands as a collection of design files combined with instructions and a community of eager builders on Discord. This relationship as a “digital” object has given the project exceptional flexibility to accommodate design improvements.
Jubilee is intended to be a small-scale multi-tool cnc platform that readily lends itself to creating custom machine-driven workflows. These often take the form of new tools and/or custom bed plates in Jubilee’s small but growing library of hardware extensions. We would like to extend this library to include new tools including a camera inspection tool, a pipetting tool, and improved workflows for liquid handling and cnc engraving. Success isnt just having a new tool or bed plate design, but also having a documented path such that anyone can reproduce workflows that use these designs.Jubilee is also meant to be friendly to newcomers while providing the reliability of a well tested platform. As such, success can also be measured by the growing number of community replications of the machine. We have currently tracked 157 people actively building Jubilees in the wild as counted by self-announcement and images posted in Jubilee’s Discord Community. We would like to grow the number of Jubilees to 250-300 in the following year.
Finally, as Jubilee is intended to be hacker friendly, enabling modifications even for new machine builders, we would like to vastly overhaul the electronics assembly instructions of the project to produce a granular experience that does not require prior machine-building experience. Through transparency in the assembly process, we are convinced that curious machine builders will more seriously consider Jubilee as a reference design towards exploring custom cnc-driven workflows, even if building Jubilee is their first foray into working with mechanical hardware.
The combined community of scientists, researchers, hackers, as well as newcomers to the maker community can benefit from a project like Jubilee.
First, the openness of the project enables people to refer to it as a reference design for building their own small scale tool changing cnc platforms. The ability to scrutinize the design from the cad model directly, to the bill-of-materials and step-by-step instructions provides community-vetted design patterns for building similar platforms.
Second, for those building Jubilees or simply curious about the project, they can take part in a rich conversation about machine building in Jubilee’ Discord server. Here they can take advantage of the community “hive mind” in a newcomer-friendly space of people eager to build and extend the design as well as patiently troubleshoot other people’s setups. The welcoming nature of the community has spurred many new builders to comment that it surpasses the utility of a forum. It’s also a place to seek inspiration for new applications to share insights and to celebrate victories.
Finally the tool library provides a small but growing number of almost ready-to-go applications for various uses including milling, multi-material 3d printing, gel extrusion, optical inspection, and rudimentary liquid handling. The breadth of applications makes Jubilee a serious consideration for anyone looking for custom automation applications and who doesn’t want to start from scratch.