Top 5 Tracking Dashboards
Each winner was awarded a Dashboards Prizes ($500 value)
Top 5 Machine Learning Models
Each winner was awarded a Machine Learning Prizes ($500 value)
Overview
In response to the growing crisis facing Africa’s diminishing elephant population, Hackster.io and Smart Parks are coming together with leading technology and conservation partners to protect the gentle giants in their natural habitats.
Elephant deaths and abuse like those pictured below are much too common, but they can be stamped out with stronger legislation, law enforcement, and conservation. In order to make that a reality, pro-conservation teams need to better data on what is happening, something our collaboration and technology are perfectly positioned to provide.
At current rates, species extinction is a possibility in 10 years' time.
We're inviting all Hacksters to join us as we develop the world's most advanced elephant tracking collar, which can help park rangers reduce animal loss from illegal ivory poaching, trophy hunting, human conflict, and environmental degradation. With #ElephantEdge, we're calling on the community to build ML models using the Edge Impulse Studio and tracking dashboards using Avnet's IoTConnect, which will be deployed onto 10 production-grade collars manufactured by our engineering partner, Institute IRNAS, deployed by Smart Parks.
All 10 winners will have an opportunity to showcase their projects on Microsoft IoT Developers’ Project 15 YouTube Channel in an ElephantEdge feature deep-dive video. We can't wait to see your Edge Impulse models and Avnet IoT Dashboards!
This will be a great way to reach the large developer community in Microsoft IoT’s leading enablement platform! Share your projects and contribute to driving awareness of ways we can continue to help the Elephant population.
Microsoft’s Project 15 is a community and an Open Platform on GitHub to help bring our Internet of Things (IoT), Artificial Intelligence (AI), and Cloud technologies to the problems facing researchers, conservationists, and others fighting to save species and habitats from extinction aka.ms/MSIoTDevs
Building machine learning models:Build ML models with Edge Impulse that will be used for collar deployments. These new models will create a new Human to Elephant Language, powered by TinyML, that will help conservation efforts:
- Poaching Risk Monitoring: Build models that can identify an increased risk for poaching. Example: Know when an elephant is moving into a high-risk area and send real-time notifications to park rangers.
- Human Conflict Monitoring: Build models and dashboards that can prevent conflict between humans and elephants. Example: Sense and alert when an elephant is heading into an area where farmers live. This collar could detect human presence by scanning if any mobile phones or WiFi hotspots are near, by tapping the available radio modules (Nordic Semiconductor nRF52840, NINA-B30x BLE, Semtech LR1110).
- Elephant Musth Monitoring: Build models and dashboards that can recognize when an elephant bull is in musth (a periodic condition in male elephants characterized by highly aggressive behavior and accompanied by a large rise in reproductive hormones). Example: Utilize the motion and acoustic sensors to discern this state of erratic, loud, and aggressive behavior.
- Elephant Activity Monitoring: Build models and dashboards that can classify the general behavior of the elephant, such as when it is drinking, eating, sleeping, etc. Example: Detect and report the elephant activity by using accelerometer data. Or go more advanced and use a water detection sensor to see when the elephant goes swimming, drinking, or digging for water.
- Elephant Communication Monitoring: Build models and dashboards that can listen for vocal communications between elephants via the onboard microphone. Example: Use sound-recording technology to listen to their vocalizations. Here's how.
This is an urgent problem that no one has totally solved. Do you have completely out-of-the-box ideas that have never considered?
Tidbits to think about:
- Elephants are afraid of bees
- Elephants speak with their feet and have poor hearing
- Elephants radiate excess heat away from the body using their ears...
- Elephants contribute to seed distribution
- Elephants perform funerals-like rituals
- LILA BC is a repository for data sets for machine learning (ML) researchers
- Elephant Voices provides you with an Elephant call types database
- Human to Elephant language "translator" https://helloinelephant.com/
- Elephants acoustics whitepaper
- What Elephant Calls Mean: A User’s Guide
- Learning what elephants are saying by the Elephant Listening Project
- Courtesy of Alasdair Davies from Arribada.org: GitHub repo and sensor data containing thermal elephant dataset collected at the ZSL Whipsnade Zoo, made available for use by anyone wishing to train their own models using Edge Impulse for the ElephantEdge competition ( (GPLv3).
Build an IoTConnect dashboard that will be used for collar deployments and help park rangers, track, monitor, and get on-demand alerts that are critical to conservation efforts:
- Simulate dashboards that track location and leaving protected areas
- Build dashboards that report the frequency of entering high-risk areas
- Monitor and infer active period vs resting period for the elephants
- Simulate alerts when activity deviates from the expected range
- Alerts to replace batteries or when a collar malfunctions, goes offline
- Design and ideate any other telemetry data and inference
The new collar will use the following hardware and software:
- Nordic Semiconductor nRF52840 Bluetooth 5, Thread, and Zigbee multi-protocol SoC.
- u-blox NINA-B30x BLE module and ZOE-M8G GPS/GNSS module.
- Taoglas low-power wide-area (LPWA) antennas.
- Western Digital Edge SDSDQAB-016G microSD storage.
- Semtech LR1110 ultra-low-power LoRaWAN transceiver.
- STMicroelectronics LIS2DW12 ultra-low-power high-performance three-axis MEMS accelerometer, with configurable single/double-tap recognition, free-fall, wakeup, portrait/landscape, and 6D/4D orientation detections.
- STMicroelectronics LSM303AGR ultra-compact high-performance eCompass module, with a 3D digital linear acceleration sensor and a 3D digital magnetic sensor.
- STMicroelectronics MP23ABS1 high-performance MEMS audio sensor with single-ended analog bottom-port microphone.
- IoTConnect Platform for dashboard creation of asset tracking.
- Edge Impulse Studio TinyML modeling software.
You do not need any hardware to build the ML models. Use datasets to sample, analyze, and build your TinyML models. You can also use your mobile phone to run simulated data collections and deployment.
You do not need specific hardware to build the dashboards. Use any hardware you already have, from Arduino to Microchip, Seeed, Adafruit to STMicroelectronics, to send data to the IoTConnect platform.
Final software and hardware will be documented and shared freely under the open source license of opencollar.io. Thank you for participating!
Prizes
We're awarding the top ten most relevant and ready-to-use projects. All projects submitted will be showcased on our partners' platform hubs after the contest.
Top 5 Tracking Dashboards
The top 5 dashboards that are relevant and ready to use will win an Apple Watch 3, a custom-ordered collectible t-shirt, and an opportunity to showcase your project on Microsoft IoT Developers’ Project 15 YouTube Channel in an ElephantEdge feature deep-dive video!
Apple Watch 3, custom tee, interview with The IoT Show by Microsoft. Must use IoTConnect.
Top 5 Machine Learning Models
The top 5 models that are relevant and ready to use will win an Apple Watch 3, a custom-ordered collectible t-shirt, and an opportunity to showcase your project on Microsoft IoT Developers’ Project 15 YouTube Channel in an ElephantEdge feature deep-dive video!
Apple Watch 3, custom tee, interview with The IoT Show by Microsoft. Must use EdgeImpulse
Judges
- Sarah Maston: Senior Solution Architect at Microsoft, Founder of Project 15
- Eric Becker: World Wildlife Fund, Conservation Program, Senior Conservation Engineer
- Grace Chen: Marketing Manager at Western Digital
- Zach Shelby: Founder and CEO at Edge Impulse
- Tim van Dam: Founder, Smart Parks
- Michael Ammann: Vice-President Platform Partnerships at u-blox
- Dan Morris: Principal Scientist, Program Director, Microsoft AI for Earth
- Vibhu Bhutani: Chief Strategy Officer, Softweb, An Avnet Company
Resources
- Want to speak "Elephant?" A new website helps you translate human words and emotions into a form of elephant communication.
- Identification of behaviors from accelerometer data in a wild social primate
- Acoustic early warning system for endangered rhinos
- Elephant listening project
- Edge Impulse TinyML Studio
- Project 15 from Microsoft: An animal conservation initiative
- Microsoft AI for Earth
- Automated Elephant Entry Prevention for Human and Crop Protection
- First Smart Parks Elephant Collar successfully deployed in Liwonde
Edge Impulse enables developers to create the next generation of intelligent device solutions with embedded Machine Learning.
The data forwarder is used to easily relay data from any device to Edge Impulse over serial. Devices write sensor values over a serial connection, and the data forwarder collects the data, signs the data, and sends the data to the ingestion service.
The Edge Impulse Uploader signs local files and uploads them to the ingestion service. This is useful to upload existing data sets, or to migrate data between Edge Impulse instances.
How much data do I need to capture?
- Accelerometer data: It is recommended to capture at least 3 minutes of data per type of movement. By default, we use a 62.5 Hz sampling frequency for accelerometers but you are free to use a different frequency as long as all your samples use the same value.
- You can capture raw data from your sensor and then apply signal processing algorithms using our DSP blocks.
- Project example: https://docs.edgeimpulse.com/docs/continuous-motion-recognition
- Microphone data: Try to get a minimum of 5 minutes for each audio category you wish to classify. We recommend recording a mono / 16 bit / 16 kHz sampling signal.
- Project example: https://docs.edgeimpulse.com/docs/audio-classification
- Try to capture as diverse data as possible to be representative of real-world conditions in the field. You may want to add some "real world" noise to increase the robustness of your model.
Which value should I set for window size and window increase?
Before applying signal processing or machine learning algorithms, your dataset will be cut in multiple windows. A sliding window mechanism is used to go through your entire raw data.
Window size represents a single data sample, it is the input of the following DSP block. The window should basically contain a full movement or a full audio extract required to distinguish the class. For instance, to detect footsteps, a 2-second window seems enough to make the distinction with other sounds.
Window increase represents the increased value of the sliding window in your raw data. For instance, imagine you have a 5-second raw data, a 1-second window size, and a window increase of 200ms, your first windows will be as follows:
Window 1: 0-1000ms
- Window 1: 0-1000ms
Window 2: 200-1200ms
- Window 2: 200-1200ms
Window 3: 400-1400ms
- Window 3: 400-1400ms
Having a low window increase value will generate more data samples for your impulse. On the other end, too many windows might overfit your Neural Network as samples could get too similar to each other.
How can I send data to my Edge Impulse project?
You have different ways to send data to your project:
- Ingestion API: For data already stored locally. Encode your data in JSON or CBOR format and send it using our Ingestion API.
- Ingestion API: For data already stored locally. Encode your data in JSON or CBOR format and send it using our Ingestion API.
- CLI Data Forwarder: Install our CLI to stream data samples directly from your device. The CLI listens to a serial port and forwards the data to your Edge Impulse project
- CLI Data Forwarder: Install our CLI to stream data samples directly from your device. The CLI listens to a serial port and forwards the data to your Edge Impulse project
- Edge Impulse Uploader: Use the UI to upload your data samples stored locally. The Uploader supports: .json, .cbor, .wav and .jpg formats
- Edge Impulse Uploader: Use the UI to upload your data samples stored locally. The Uploader supports: .json, .cbor, .wav and .jpg formats
- Cropping Items: Ever captured too much data? There's an easy way to trim it now.
What if my model performs badly?
Check out the Increasing model performance guide.
You can also ask your question on the Edge Impulse forum.
IoTConnect
IoTConnect is not merely a system that allows you to manage connected assets. Our engineers have carefully chosen the state-of-the-art technologies to build the most advanced and high-performing IoT platform that helps you connect things quickly, build a data-driven business, generate new insights, and drive actions. IoTConnect is a Platform as a Service (PaaS) that facilitates device communication and management, data storage, and app creation while adhering to robust security protocols.
What can IoTConnect do?
IoTConnect is a full-fledged Platform as a Service (PaaS). This horizontal IoT platform allows for device communication and management, data storage, app creation and enablement, robust security protocols, and implementation of data science methodologies.
Sample dashboards
https://www.iotconnect.io/remote-asset-monitoring-solutions.html
Sample case study
Get access & login
IoTConnect access can be provided by sending an email to elephantedge@iotconnect.io. You will receive an invitation link to an easy and free set up in 2 business days. Log in with your credentials at avnet.iotconnect.io and look for your unique company id from the company profile page. Your IoTConnect license will last 60 days.
All the participants that request account will get access in 2 business days and the account will have the following restrictions:
- Number of messages per month – 50k ( around 1500 per day ). Auto Stop after 50k messages
- Monthly subscription (30 days) which will be auto renewed for 3 months (3 Times, Will be stopped after 3 renewals)
- Max Devices: 3
- Max Users: 2
Basic concepts IoTConnect SDK’s, Data flow, SDK’s Public Methods
SDK solves the purpose of D2C and C2D communications, SDK works as a mediator between device and cloud platforms. IoTConnect Device SDKs include a set of tools, libraries, developer guides with code samples, and porting guides to help you in quickly and easily connect your devices to IoTConnect.
Choose your SDK or build your own & connect your devices
You can choose your desired SDK or build your own if one if any of the SDK is not right for the use case. Once the SDK is downloaded follow the steps to connect your data to IoTConnect.
Contest Status
Timeline
Competition begins
August 11, 2020 at 5:00 AM PDT
Submissions close
October 30, 2020 at 11:59 PM PDT
Winners announced by
Nov 18, 2020