Algae Attack

Inspiration

As a group that is predominantly from Florida (Go Gators!) and the coast, we have grown up surrounded by water out whole lives. Recently however, our homes have been more and more put at risk due to the increased frequency of Red Tide.

Red Tide is a phenomenon that occurs when algae grows in a body of water very rapidly and causes the water to turn red via the production of toxins. These toxins can cause burning and irritation of the eyes and skin in humans. In aquatic life, these toxins can greatly harm and cause a massive loss of life due to the effects of the toxin on marine animal.

The effects on humans while problematic is nothing compared to the massive loss of aquatic life that occurs within our communities. One reason why these issues keep occurring is because the agencies that typically attack these issues are often chronically underfunded so if this issue is going to be solved, a smarter solution is needed. That is where we come in

What it does

We have developed Algae Attack as a proof of concept device that can be used to make the monitoring, tracking, and response of Algae blooms much more manageable and cheaper. To start off we have written 2 major sets of code. One of these sets of code is simulating the field IOT devices that would go in the wild and measure the parameters of water which is often done by an environmental scientist. This IOT device sends important metrics to a web-server we have hosted in the Cloud where it will be stored. This data is then fed forward into our Machine Learning algorithm where it predicts the likelihood that an Algae bloom will occur. This data is then passed to our front end where it is integrated so users can easily track and see the severity levels of future algal blooms!

How we built it

Using the MERN Web-stack, TensorFlow, and Python

Challenges we ran into

Integrating all parts of the project together proved to be challenging but luckily after a very sleepless night we pulled through.

Accomplishments that we're proud of

Creating a full stack Web-app, machine learning algorithm, and functional and reasonable manufacturing plan.

What we learned

How to work and code without sleep 😜. We also gained knowledge in the domain knowledge of this issue as well as preprocessing the feature and image datasets. In order to better determine which features to use for our model, we read a few research papers that tested various models on ecological data. We learned how harmful algae blooms (hab) form (from excess phosporus and nitrogen in bodies of water with algae), the harmful effects of habs, and the chemistry behind the how various features and materials affect the formation of and deformation of habs. We also gained knowledge in many ways to implement n-way-k-shot learning models. Some n-way-k-shot learning models that we analyzed include siamese networks, triplet networks, multi-class comparators, matching networks, and prototypical networks.

What's next for Algae Attack

We plan on improving our app with more functionality. One that we were working on but didn't quite complete was to utilize an N-way-K-shot classification neural network to classify pictures of bodies of water with a very limited training set to determine the likelihood of the body of water being affected by a harmful algae bloom (hab). After that we plan on looking into more computer vision algorithms to classify the video footage from the camera that our IOT device will retrive to hopfully create a more accurate model for determining if a body of water is effected by a hab. In addition, we hope to make the IOT device more robot-like by adding in thrusters with remote-control functionality so the IOT device can collect data from various parts of the lake.

Built With

• mern
• python
• tensorflow