Inspiration

In this Climate Hackathon, we chose to help the organization Buy Food With Plastic to automate their workflow in plastic waste collection.

Arun: Back In India, I use to visit Marina beach every weekend, to witness the beautiful sunset from the shores of the second longest beach in the world. It's very disheartening to see, How Marina beach has transformed over the years, hiding its eternal beauty under the shades of plastics. One can witness a lot of slums near Marina, with tons of plastic wastes thrown over the shores near-by without any awareness about the perils of plastics scattered over the shores. The kind of wastes spread across the shore has really taken a toll on the number of people visiting the beach. Yeah, Even I felt weird to sit at some parts of the smelly beaches which has brought me more stress rather than peace. At the same time, You can see quite a number of people from slums struggling to sell some random products for a very small amount of price, just with the hope to feed their family at least once a day. Definitely, It's agreeable that, When people don't have money to feed their family, We can't expect them to show awareness towards the environmental factors. When we discussed the potential hackathon idea, this one idea about Buy Food With Plastic stood out and brought back my memories of my hometown. If some automated system can provide food for the collected waste, then this is a sure-shot project that can provide a win-win situation for the people struggling in the slums of Marina. This was the basic thought that triggered us to glue our focus towards this project.

What it does

Eruza is a Waste Automation & Management System that automates the end-to-end process of waste collection and aims to consolidate all the data required for the generation of impact report in one place.

Using our mobile web application, the volunteers can quickly capture image or voice content which is analyzed and transformed to metrics that can be included in their report with ease. The solution makes it easy to run waste collection efforts, letting organizations focus on the social and environmental impact.

How we built it

Azure Solution

The solution architecture was drafted as seen in the figure below. From there, resources in the Azure portal were added accordingly. That is, Along with that, an ARM (Azure Resource Management) template project was created to ensure reproducibility.

Azure Resources used were:

├── Cognitive Services
├──── Computer Vision
├──── Speech
├──── Custom Vision
├── Azure Storage
├──── Blob Storage (static web + application data)
├──── 
├── Container Registry + IoT Central (for the IoT Edge experiment)

Azure system architecture

Web Application

A progressive web app, connected with Azure resources via Azure SDK for JavaScript. As a starting point, we used the Azure-Sample repo AzureSpeechReactSample

app screenshots

Example use case

Let's assume the organization team brings small scales to measure the weight of bags filled with plastic bottles. This one in the image below contained 8 big bottles = 0.45. That is, if we assume one bottle is ~55 grams, we could estimate this bag contains 8 bottles. The user can open our app and make a voice record "450 grams" and pick the "Bottles" option. From there, it's possible to save the data and use it in the Impact Report.

app screenshots

IOT EDGE

Using Azure IoT Edge, we distributed our object detection model to a Raspberry Pi. On the device, Docker and the IoT Edge Runtime was installed and we deployed two modules (Docker images):

  • A camera capturing module (from this Azure-Sample repo )
  • A object detection model (trained and exported from Custom Vision)

This part of the project was mainly a proof of concept, as we imagined having cameras at the waste stations. Our process was:

  • collecting a dataset
  • annotate images and train model in the Custom Vision studio
  • prepare Raspberry Pi with a connection string from IoT Central
  • create device template and deploy, then watch the telemetry data arrive

dataset

IoT Edge camera scenario

Challenges we ran into

  • Working and collaborating in different time zones
  • Getting a hold on different APIs of Azure Cognitive Services

Accomplishments that we're proud of

  • That we managed to get 3 implementations of AI capabilities that could help to automate the workflow.
  • We got together to problem solve and be critical thinkers to help our bigger cause.

What we learned

  • Working with the Azure SDK for JavaScript and especially Azure Cognitive Services
  • Teamwork and commitment towards the goal

What's next for Eruza - Waste Automation & Management System

  • Automatic generation of PDF file of Impact Report
  • Complete data uploads from web app to the database
  • From AI results, extract the relevant values. E.g. from "donation of 5 dollars from the Juarez family", extract the number 5 to be added to the total value.

  • The app can also be reused for Smart city administration, Marine clean-up, Disaster recovery, etc with slight modifications.

Share this project:

Updates