STEM-In-Action Spring Scoop: The Marvelous Mosquito Marauders

The time has come! We have reached our very last 2020 STEM-In-Action grant winner's Spring Scoop.  If you're just tuning into the eCYBER Blog, welcome. The U.S. Army Educational Outreach Program (AEOP) STEM-In-Action Grant awards eCYBERMISSION teams up to $5,000 to develop their projects into mature and scalable solutions in their community. Typically, we award this honorary grant to five teams, but in 2020 ten lucky teams took home the prestigious award. For the past few months we have caught up with all our lucky grant winning teams to see how their projects have been progressing since they were granted the prestigious award at last year's NJ&EE. Today we're catching up with our final SIAG winners, the Marvelous Mosquito Marauders!

--

Hello again! The Marvelous Mosquito Marauders have come to update you on our project since receiving eCYBERMISSION’s STEM-in-Action Grant. Just a reminder or for those just tuning in, our team consists of Ajay, Amulya, Avni, and Samvrit, 8th graders from Stone Hill Middle School located in Ashburn, Virginia. Our mission is to reduce the number of mosquito-borne diseases with ArboTrack, a crowd-sourced, geolocation, and photo-based app that accurately locates areas of standing water. This year, we are using eCYBERMISSION’s STEM-In-Action Grant to implement ArboTrack in both the local and global communities. 

With the Stem-in-Action Grant, we concentrated on three main components within our project; App Updates, Machine Learning Model, and Website/Community Outreach. Splitting our project into various groups allowed us to work on these essential elements simultaneously and allow us to give each other constructive feedback on how to achieve more within that domain. Also, since we have a limited time frame to effectively implement our solution within our community, dividing the work would allow us to accomplish our tasks in a timely manner. 

The first section of our project was making significant updates to our app by improving its efficiency and user-friendliness. We have made front-end enhancements such as adding clear buttons and text on the user interface, which will accommodate users without having difficulties when using ArboTrack. For more reliability, we made sure users were who they registered to be by making all the registration and login fields mandatory. To add an extra layer of security, city employees are now validated with a confirmation email. A major issue that we faced while testing our app was that some users would enter the latitude and longitude, which confused the Google Maps API. So, a text was added, notifying the user that they didn’t have to enter anything. 

The second section of our project was creating a Machine Learning (ML) Image Classification model to filter false images from actual reports of standing water. Our ML model is an algorithm that is trained over a dataset of standing water images and false reports to recognize patterns in the data, allowing the model to correctly sort new images that are water from spam. Our first model was created on CustomVision and didn’t need many images for a 75%+ accuracy rate. We then started creating a model using Python to better fit our needs. We were able to find some online courses to start our model. Our first step was to collect images of standing water and false reports, such as trees and roads, to use as our training and testing data, collecting around 100 images in each data set, and after some image editing, we were able to get to around 500. Since this is less than expected (more training data for the model meant higher accuracy when recognizing new images) our accuracy only reached about 85%. We were able to convert our model into an API, hosted it in the Cloud, which takes an image and outputs if it is standing water or not. Finally, we integrated the new API with an Azure Logic App, which receives images from our app database and can push the predictions back into the database. 

The final section of our project is the Community Outreach. In this section, we made a website that showcases ArboTrack, includes statistics on mosquito-borne diseases, and a blog page where we post updates/stories about our product. Our website also has a way to sign up for our newsletter! Our website can be found directly at https://arbotrack.org, or you can search for arbotrack on the search engine of your choice (it is the first result!) As of Community Outreach, we have three social media accounts, which include Facebook (ArboTrack App), LinkedIn (ArboTrack App), and Twitter (@ArboTrack), where we have posted videos and updates of our app. We also created a survey that we sent to about 50 mosquito-borne disease health officials, who live in prevalent mosquito areas. The questions the health officials would answer would help us further develop our app with their keen suggestions on how to improve it to better fit their needs. Finally, we created a webinar presentation that we will use to bring awareness to ArboTrack. 

We have encountered some issues while working on Community Outreach. Due to the pandemic, we cannot have presentations in public areas such as libraries to bring awareness to ArboTrack. We tried solving this issue by posting on our social media accounts, but not many people pay attention to accounts with few followers. We decided to have recorded webinars and post them on our website and social media accounts, as we cannot do any live presentations. Also, we tried getting in contact with local news channels and agencies, but we are currently waiting for responses. We will continue contacting news agencies and present our project, showcasing ArboTrack, in front of Mosquito Abatement Agencies and HOA members. Hopefully, we will receive feedback so we can implement their suggestions in our app.

Despite the issues we have encountered, we are sticking with our timeline and have exciting future plans and goals we are going to meet! Real-time goals for our app include continuing to enhance ArboTrack such as linking the physician to a specific area/report, which will enable a specific physician to be assigned to a specific area. We plan to continue training our ML model with more new images and then improve our model. For community outreach, we will record and post our webinar on our social media accounts so others are aware of our app. Goals in the future include publishing ArboTrack in Apple’s App Store and the Google Play Store, integrating our app with Mosquito Abatement Agency (MAA) software, and gathering feedback from citizens, physicians, city employees, and Mosquito Abatement Agency officials. The most ideal outcome of our project would be integrating our app with MAA software and receive feedback on improving it to fit MAA officials’ needs. We are planning to publish our app during Spring, but we may be held back due to time-frame constraints. We are very excited to publish ArboTrack in both the Apple/Google Play stores! We can’t wait to see various users using our app and making our community a better place to live in with the touch of a button!

--

As is typical of our STEM-In-Action grant recipients, we know this team is onto effect something bigger than just their local community. Mosquito-borne diseases are a global problem and with the progress of their ArboTrack app, their effectiveness in combatting these disease could have a significant effect on the global community. We cannot wait to see what becomes of their app in the future!

-Mission Control

Colleen Minan
AEOP Communications & Marketing Specialist
cminan@nsta.org