Project Info

Winner

Intel: Best Use of Intel Developer Cloud ($10k Credits + 4x Lenovo AI PC [1st] & $4k Credits [2nd] & $2k Credits [3rd])

Meshworks - NLP LoRa Mesh Network for Emergency Response

Inspiration

In times of disaster, the capacity of rigid networks like cell service and internet dramatically decreases at the same time demand increases as people try to get information and contact loved ones. This can lead to crippled telecom services which can significantly impact first responders in disaster struck areas, especially in dense urban environments where traditional radios don't work well. We wanted to test newer radio and AI/ML technologies to see if we could make a better solution to this problem, which led to this project.

What it does

Device nodes in the field network to each other and to the command node through LoRa to send messages, which helps increase the range and resiliency as more device nodes join. The command & control center is provided with summaries of reports coming from the field, which are visualized on the map.

How we built it

We built the local devices using Wio Terminals and LoRa modules provided by Seeed Studio; we also integrated magnetometers into the devices to provide a basic sense of direction. Whisper was used for speech-to-text with Prediction Guard for summarization, keyword extraction, and command extraction, and trained a neural network on Intel Developer Cloud to perform binary image classification to distinguish damaged and undamaged buildings.

Challenges we ran into

The limited RAM and storage of microcontrollers made it more difficult to record audio and run TinyML as we intended. Many modules, especially the LoRa and magnetometer, did not have existing libraries so these needed to be coded as well which added to the complexity of the project.

Accomplishments we're proud of

: We wrote a library so that LoRa modules can communicate with each other across long distances We integrated Intel's optimization of AI models to make efficient, effective AI models We worked together to create something that works

What we learned

: How to prompt AI models How to write drivers and libraries from scratch by reading datasheets How to use the Wio Terminal and the LoRa module

What's next

We will improve the audio quality captured by the Wio Terminal and move edge-processing of the speech-to-text to increase the transmission speed and reduce bandwidth use. We will add a high-speed LoRa network to allow for faster communication between first responders in a localized area We will integrate the microcontroller and the LoRa modules onto a single board with GPS in order to improve ease of transportation and reliability

Analysis

Compare with all teams

View

Metric

Figures cover GitHub contributors during the hackathon window. A co-authored commit counts in full for each author, so per-member totals add up to more than the whole-team figures.

Technology

Found in codeClaimed only

CSSIn code
HTMLIn code
JavaScriptIn code
PythonIn code
FlaskClaimed
PyTorchClaimed

4 of 6 appear in the indexed code. 2 claimed on Devpost could not be matched to code, which may simply mean the tool leaves no trace in the repository.

AI coding agents

No AI coding agent signals were found in this repository.

Detected from committed agent config files and commit authorship. Absence of a signal is not proof an agent was unused.

Codebase size

Source size

56 KB

Source files

Counts recognized source files only; vendored directories, binaries and lockfiles are excluded, so this is smaller than the repository on disk.